There has been considerable research on thin-slice judgments, but no study has demonstrated the predictive validity of confidence measures when assessors watch videos acquired from communication systems, in which the perceptual quality of videos could be degraded by limited bandwidth and unreliable network conditions. This paper studies the relationship between high-level thin-slice judgments of human behavior and factors that contribute to perceptual video quality. Based on a large number of subjective test results, it has been found that the confidence of a single individual present in all the videos, called speaker’s confidence (SC), could be predicted by a list of features that contribute to perceptual video quality. Two prediction models, one based on artificial neural network and the other based on a decision tree, were built to predict SC. Experimental results have shown that both prediction models can result in high correlation measures.

This PDF file contains the editorial “Special Section Guest Editorial:Color in Texture and Material Recognition” for JEI Vol. 25 Issue 06

Color matching of fabric blends is a key issue for the textile industry, mainly due to the rising need to create high-quality products for the fashion market. The process of mixing together differently colored fibers to match a desired color is usually performed by using some historical recipes, skillfully managed by company colorists. More often than desired, the first attempt in creating a blend is not satisfactory, thus requiring the experts to spend efforts in changing the recipe with a trial-and-error process. To confront this issue, a number of computer-based methods have been proposed in the last decades, roughly classified into theoretical and artificial neural network (ANN)–based approaches. Inspired by the above literature, the present paper provides a method for accurate estimation of spectrophotometric response of a textile blend composed of differently colored fibers made of different materials. In particular, the performance of the Kubelka-Munk (K-M) theory is enhanced by introducing an artificial intelligence approach to determine a more consistent value of the nonlinear function relationship between the blend and its components. Therefore, a hybrid K-M+ANN-based method capable of modeling the color mixing mechanism is devised to predict the reflectance values of a blend.

The food industry is one of the industries that uses machine vision for a nondestructive quality evaluation of the produce. These quality measuring systems and softwares are precalculated on the basis of various image-processing algorithms which generally use a particular type of classifier. These classifiers play a vital role in making the algorithms so intelligent that it can contribute its best while performing the said quality evaluations by translating the human perception into machine vision and hence machine learning. The crop of interest is rice, and the color of this crop indicates the health status of the plant. An enormous number of classifiers are available to solve the purpose of color prediction, but choosing the best among them is the focus of this paper. Performance of a total of 60 classifiers has been analyzed from the application point of view, and the results have been discussed. The motivation comes from the idea of providing a set of classifiers with excellent performance and implementing them on a single algorithm for the improvement of machine vision learning and, hence, associated applications.

Texture description is a challenging problem with color images. Despite some attempts to include colors in local binary patterns (LBPs), no proposal has emerged as a color counterpart of grayscale LBPs. This is because colors are defined by vectors that are not naturally ordered and several ways exist to compare them. We propose an LBP extension that takes the vector information of color into account due to a color order. As several color orders are available and the selection of the most suitable one is difficult, we combine two of them in a texture descriptor called “mixed color order LBPs.” This small-size feature provides good performance on several benchmark databases for two classification problems with regard to larger-size LBP-based features of color textures.

Considering that remote sensing images contain rich scale-dependent information and geographical detailed information, segmentation process must be carried out under the multiscale case. The vector-valued C-V active contour model is an effective image segmentation method, but the model cannot segment the nonhomogeneous remote sensing images well. The image processing methods based on nonsubsampled contourlet transform (NSCT) can fully use the detailed information of remote sensing images. The interscale distribution characteristics of NSCT coefficients at finer scale is first analyzed and then a statistical model of signal singularities combining the coefficient correlation between intrascale and interscale is proposed. Based on the above, the vector-valued C-V active contour model is then applied to the statistical characteristics for segmenting images. Consequently, the proposed method can preserve detailed information of images and other desirable properties of active contour model. Numerical examples indicate that the proposed method is very competitive with several state-of-the-art techniques.

This paper presents a comparison of color spaces for material classification. The study includes three device-independent (CIELAB, CIELUV, and CIE XYZ) and seven device-dependent spaces (RGB, HSV, YIQ, YUV, YC_bC_r, Ohta’s I₁I₂I₃, and RG-YeB-WhBl). The pros and cons of the different spaces and the procedures for converting color data among them are discussed in detail. An experiment based on 12 different image data sets was carried out to comparatively evaluate the performance of each space for material classification purposes. The results showed that CIELAB markedly outperformed the other spaces followed by HSV and CIELUV. Conversely, CIE XYZ came out as the worst performing space. Interestingly, no significant difference emerged among the performance of the other device-dependent spaces.

Facial expression recognition in the wild is a very challenging task. We describe our work in static and continuous facial expression recognition in the wild. We evaluate the recognition results of gray deep features and color deep features, and explore the fusion of multimodal texture features. For the continuous facial expression recognition, we design two temporal–spatial dense scale-invariant feature transform (SIFT) features and combine multimodal features to recognize expression from image sequences. For the static facial expression recognition based on video frames, we extract dense SIFT and some deep convolutional neural network (CNN) features, including our proposed CNN architecture. We train linear support vector machine and partial least squares classifiers for those kinds of features on the static facial expression in the wild (SFEW) and acted facial expression in the wild (AFEW) dataset, and we propose a fusion network to combine all the extracted features at decision level. The final achievement we gained is 56.32% on the SFEW testing set and 50.67% on the AFEW validation set, which are much better than the baseline recognition rates of 35.96% and 36.08%.

We adopt genetic programming (GP) to define a measure that can predict complexity perception of texture images. We perform psychophysical experiments on three different datasets to collect data on the perceived complexity. The subjective data are used for training, validation, and test of the proposed measure. These data are also used to evaluate several possible candidate measures of texture complexity related to both low level and high level image features. We select four of them (namely roughness, number of regions, chroma variance, and memorability) to be combined in a GP framework. This approach allows a nonlinear combination of the measures and could give hints on how the related image features interact in complexity perception. The proposed complexity measure MGP exhibits Pearson correlation coefficients of 0.890 on the training set, 0.728 on the validation set, and 0.724 on the test set. MGP outperforms each of all the single measures considered. From the statistical analysis of different GP candidate solutions, we found that the roughness measure evaluated on the gray level image is the most dominant one, followed by the memorability, the number of regions, and finally the chroma variance.

Color-band resistor possessing specular surface is worthy of studying in the area of color image processing and color material recognition. The specular reflection and halo effects appearing in the acquired resistor image will result in the difficulty of color band extraction and recognition. A computer vision system is proposed to detect the resistor orientation, segment the resistor’s main body, extract and identify the color bands, as well as recognize the color code sequence and read the resistor value. The effectiveness of reducing the specular reflection and halo effects are confirmed by several cheap covers, e.g., paper bowl, cup, or box inside pasted with white paper combining with a ring-type LED controlled automatically by the detected resistor orientation. The calibration of the microscope used to acquire the resistor image is described and the proper environmental light intensity is suggested. Experiments are evaluated by 200 4-band and 200 5-band resistors comprising 12 colors used on color-band resistors and show the 90% above correct rate of reading resistor. The performances reported by the failed number of horizontal alignment, color band extraction, color identification, as well as color code sequence flip over checking confirm the feasibility of the presented approach.

The analysis of color and texture has a long history in image analysis and computer vision. These two properties are often considered as independent, even though they are strongly related in images of natural objects and materials. Correlation between color and texture information is especially relevant in the case of variable illumination, a condition that has a crucial impact on the effectiveness of most visual descriptors. We propose an ensemble of hand-crafted image descriptors designed to capture different aspects of color textures. We show that the use of these descriptors in a multiple classifiers framework makes it possible to achieve a very high classification accuracy in classifying texture images acquired under different lighting conditions. A powerful alternative to hand-crafted descriptors is represented by features obtained with deep learning methods. We also show how the proposed combining strategy hand-crafted and convolutional neural networks features can be used together to further improve the classification accuracy. Experimental results on a food database (raw food texture) demonstrate the effectiveness of the proposed strategy.

We present the implementation of a feature extraction approach for the automated screening of Salmonella sp., a task visually carried out by a microbiologist, where the resulting color characteristics of the culture media plate indicate the presence of this strain. The screening of Salmonella sp. is based on the inoculation and incubation of a sample on an agar plate, allowing the isolation of this strain, if present. This process uses three media: Xylose lysine deoxycholate, Salmonella Shigella, and Brilliant Green agar plates, which exhibit specific color characteristics over the colonies and over the surrounding medium for a presumed positive interpretation. Under a controlled illumination environment, images of plates are captured and the characteristics found over each agar are processed separately. Each agar is analyzed using statistical descriptors for texture, to determine the presence of colonies, followed by the extraction of color features. A comparison among the color features seen over the three media, according to the FDA Bacteriological Analytical Manual, determines the presence of Salmonella sp. on a given sample. The implemented process proves that the task addressed can be accomplished under an image processing approach, leading to the future validation and automation of additional screening processes.

We propose a method for solving one of the significant open issues in computer vision: material recognition. A time-of-flight range camera has been employed to analyze the characteristics of different materials. Starting from the information returned by the depth sensor, different features of interest have been extracted using transforms such as Fourier, discrete cosine, Hilbert, chirp-z, and Karhunen–Loève. Such features have been used to build a training and a validation set useful to feed a classifier (J48) able to accomplish the material recognition step. The effectiveness of the proposed methodology has been experimentally tested. Good predictive accuracies of materials have been obtained. Moreover, experiments have shown that the combination of multiple transforms increases the robustness and reliability of the computed features, although the shutter value can heavily affect the prediction rates.

Image oversegmentation creates small, compact, and irregularly shaped regions subject to further clustering. Consideration of texture characteristics can improve the resulting quality of the clustering process. Existing methods based on an orthogonal transform into frequency domain can extract texture features of arbitrarily shaped regions only from inscribed rectangles. We propose a method for extracting texture features of entire arbitrarily shaped image regions using orthogonal transforms. Furthermore, we introduce a mathematically correct method for unifying spectral dimensions that is necessary for accurate comparison and classification of spectra with different dimensions. The proposed method is particularly suitable for classifying areas with periodic and quasiperiodic textures. Our approach exploits the texture periodification property of certain orthogonal transforms that is based on insertion of zeros into the spectrum. We identified some of those orthogonal transforms which possess this important property and also provide mathematical proofs of our claims. Last, we show that inclusion of luminance and chrominance components into the feature vector increases the precision of the proposed method which then becomes suitable for natural scene images as well.

This PDF file contains the editorial “Special Section Guest Editorial:Perceptually Driven Visual Information Analysis” for JEI Vol. 25 Issue 06

This article [J. Electron. Imaging 25(6), 061602 (2016), doi: 10.1117/1.JEI.25.6.061602] was retracted on 18 December 2018 due to double publication in this and another peer-reviewed journal. The authors regret this mistake.

On the basis of the different strengths of synaptic connections between actual neurons, this paper proposes a heterogeneous pulse coupled neural network (HPCNN) algorithm to perform quantization on images. HPCNNs are developed from traditional pulse coupled neural network (PCNN) models, which have different parameters corresponding to different image regions. This allows pixels of different gray levels to be classified broadly into two categories: background regional and object regional. Moreover, an HPCNN also satisfies human visual characteristics. The parameters of the HPCNN model are calculated automatically according to these categories, and quantized results will be optimal and more suitable for humans to observe. At the same time, the experimental results of natural images from the standard image library show the validity and efficiency of our proposed quantization method.

Assessment of visual quality plays a crucial role in modeling, implementation, and optimization of image- and video-processing applications. The image quality assessment (IQA) techniques basically extract features from the images to generate objective scores. Feature-based IQA methods generally consist of two complementary phases: (1) feature extraction and (2) feature pooling. For feature extraction in the IQA framework, various algorithms have been used and recently, the two-dimensional (2-D) mel-cepstrum (2-DMC) feature extraction scheme has provided promising results in a feature-based IQA framework. However, the 2-DMC feature extraction scheme completely loses image-phase information that may contain high-frequency characteristics and important structural components of the image. In this work, “2-D complex mel-cepstrum” is proposed for feature extraction in an IQA framework. The method tries to integrate Fourier transform phase information into the 2-DMC, which was shown to be an efficient feature extraction scheme for assessment of image quality. Support vector regression is used for feature pooling that provides mapping between the proposed features and the subjective scores. Experimental results show that the proposed technique obtains promising results for the IQA problem by making use of the image-phase information.

Age and gender classification are two important problems that recently gained popularity in the research community, due to their wide range of applications. Research has shown that both age and gender information are encoded in the face shape and texture, hence the active appearance model (AAM), a statistical model that captures shape and texture variations, has been one of the most widely used feature extraction techniques for the aforementioned problems. However, AAM suffers from some drawbacks, especially when used for classification. This is primarily because principal component analysis (PCA), which is at the core of the model, works in an unsupervised manner, i.e., PCA dimensionality reduction does not take into account how the predictor variables relate to the response (class labels). Rather, it explores only the underlying structure of the predictor variables, thus, it is no surprise if PCA discards valuable parts of the data that represent discriminatory features. Toward this end, we propose a supervised appearance model (sAM) that improves on AAM by replacing PCA with partial least-squares regression. This feature extraction technique is then used for the problems of age and gender classification. Our experiments show that sAM has better predictive power than the conventional AAM.

We present an approach to figure/ground organization using mirror symmetry as a general purpose and biologically motivated prior. Psychophysical evidence suggests that the human visual system makes use of symmetry in producing three-dimensional (3-D) percepts of objects. 3-D symmetry aids in scene organization because (i) almost all objects exhibit symmetry, and (ii) configurations of objects are not likely to be symmetric unless they share some additional relationship. No general purpose approach is known for solving 3-D symmetry correspondence in two-dimensional (2-D) camera images, because few invariants exist. Therefore, we present a general purpose method for finding 3-D symmetry correspondence by pairing the problem with the two-view geometry of the binocular correspondence problem. Mirror symmetry is a spatially global property that is not likely to be lost in the spatially local noise of binocular depth maps. We tested our approach on a corpus of 180 images collected indoors with a stereo camera system. K-means clustering was used as a baseline for comparison. The informative nature of the symmetry prior makes it possible to cluster data without a priori knowledge of which objects may appear in the scene, and without knowing how many objects there are in the scene.

Most image processing and computer vision applications require edge detection for object recognition, image segmentation, and scene analysis. The traditional algorithms cannot handle the demanding requirements on the accuracy and robustness of these applications. Information set theory is utilized in this paper for defining edge strength measures which help in finding robust edges. The proposed work is originated from the smallest univalue segment assimilating nucleus concept, wherein a mask is applied on the red, green, and blue components of the color image for calculating a small area of neighboring pixels with similar brightness to center pixels. A symmetric Gaussian membership function (MF) is used to fuzzify the histogram of this area. This MF is converted into sigmoidal MF to strengthen and sharpen the weak edges. These two MFs provide the best results in comparison to other MFs used in literature. Extensive simulation results show that the proposed technique produces better results than other existing techniques in terms of the qualitative and quantitative measures, which include Pratt’s figure of merit, structural similarity index, and analysis of variance. The proposed technique also works well in the presence of impulse noise.

Abnormal event detection plays a critical role for intelligent video surveillance, and detection in crowded scenes is a challenging but more practical task. We present an abnormal event detection method for crowded video. Region-wise modeling is proposed to address the inconsistent detected motion of the same object due to different depths of field. Comparing to traditional block-wise modeling, the region-wise method not only can reduce heavily the number of models to be built but also can enrich the samples for training the normal events model. In order to reduce the computational burden and make the region-based anomaly detection feasible, a saliency detection technique is adopted in this paper. By identifying the salient parts of the image sequences, the irrelevant blocks are ignored, which removes the disturbance and improves the detection performance further. Experiments on the benchmark dataset and comparisons with the state-of-the-art algorithms validate the advantages of the proposed method.

This work was conducted in the context of perceptually realistic image simulations of paint materials. These simulations aim at producing the same visual response in an observer as real objects given certain illumination and observation conditions. The colorimetric validation ensures that the colors in the simulated images differ from the real scene below a given threshold and human observers cannot discern between them. Most existing validations use the direct approach of comparing renderings with real scenes and are usually briefly described. We propose a colorimetric validation using radiometrically calibrated photographs as valid perceptual references of the real scene, which avoids taking spectral measurements of the scene. This paper details important methodological aspects, providing objective validation results for a Macbeth color chart and a sample paint coating. Our results show that we can achieve xy chromaticity differences under 3%, relative luminance differences under 5%, and CIEDE2000 differences under 10% for the Macbeth chart. In the case of a sample paint coating, achieved chromaticity differences are under 11% and average luminance and CIEDE2000 differences are under 1%. Validation by visual inspection is not addressed here.

Visual information is very important in human perceiving of the surrounding world. During the observation of the considered scene, some image parts are more salient than others. This fact is conventionally addressed using the regions of interest approach. We are presenting an approach that captures the saliency information per pixel basis using one continuous saliency map for a whole image and which is directly used in the lossy image compression algorithm. Although for the encoding/decoding part of the algorithm, the notion region is not necessary anymore; the resulting method can, due to its nature, efficiently emulate large amounts of regions of interest with various significance. We provide reference implementation of this approach based on the set partitioning in hierarchical trees (SPIHT) algorithm and show that the proposed method is effective and has potential to achieve significantly better results in comparison to the original SPIHT algorithm. The approach is not limited to SPIHT algorithm and can be coupled with, e.g., JPEG 2000 as well.

Existing stereoscopic image quality assessment (SIQA) methods are mostly based on the luminance information, in which color information is not sufficiently considered. Actually, color is part of the important factors that affect human visual perception, and nonnegative matrix factorization (NMF) and manifold learning are in line with human visual perception. We propose an SIQA method based on learning binocular manifold color visual properties. To be more specific, in the training phase, a feature detector is created based on NMF with manifold regularization by considering color information, which not only allows parts-based manifold representation of an image, but also manifests localized color visual properties. In the quality estimation phase, visually important regions are selected by considering different human visual attention, and feature vectors are extracted by using the feature detector. Then the feature similarity index is calculated and the parts-based manifold color feature energy (PMCFE) for each view is defined based on the color feature vectors. The final quality score is obtained by considering a binocular combination based on PMCFE. The experimental results on LIVE I and LIVE Π 3-D IQA databases demonstrate that the proposed method can achieve much higher consistency with subjective evaluations than the state-of-the-art SIQA methods.

This paper presents an algorithm for automatic segmentation of moving objects in video based on spatiotemporal visual saliency and an active contour model. Our algorithm exploits the visual saliency and motion information to build a spatiotemporal visual saliency map used to extract a moving region of interest. This region is used to automatically provide the seeds for the convex active contour (CAC) model to segment the moving object accurately. The experiments show a good performance of our algorithm for moving object segmentation in video without user interaction, especially on the SegTrack dataset.

Research on video quality assessment (VQA) plays a crucial role in improving the efficiency of video coding and the performance of video processing. It is well acknowledged that the motion energy model generates motion energy responses in a middle temporal area by simulating the receptive field of neurons in V1 for the motion perception of the human visual system. Motivated by the biological evidence for the visual motion perception, a VQA method is proposed in this paper, which comprises the motion perception quality index and the spatial index. To be more specific, the motion energy model is applied to evaluate the temporal distortion severity of each frequency component generated from the difference of Gaussian filter bank, which produces the motion perception quality index, and the gradient similarity measure is used to evaluate the spatial distortion of the video sequence to get the spatial quality index. The experimental results of the LIVE, CSIQ, and IVP video databases demonstrate that the random forests regression technique trained by the generated quality indices is highly correspondent to human visual perception and has many significant improvements than comparable well-performing methods. The proposed method has higher consistency with subjective perception and higher generalization capability.

The problem of eliminating character-resembling blobs on a detected region in the plate detection stage of an automated license plate recognition system is addressed. The proposed method amplifies the slight differences between the noncharacter blobs (anomalies) and the character blobs (true signal) to enhance the tractability. This method postulates on two propositions: (1) the anomalies are usually located around the true signal and the suspected anomalies and (2) blobs should be given less emphasis in computing a reference point. The first proposition is based on prior knowledge and observation; the second proposition is based on the fact that a reference point that takes anomalies into account is contaminated and thus misleading. The gist of the method mainly focuses on the methodology to emphasize the blobs differently in accordance to their location in computing the reference point that approximates the representative value of true signal properties more accurately, thus giving the effect of amplifying the slight differences. The performance of the method is evaluated on both its capability and consistency in solving certain types of anomalies.

A particular algorithm for moving object detection using a background subtraction approach is proposed. We generate the background model by combining quad-tree decomposition with entropy theory. In general, many background subtraction approaches are sensitive to sudden illumination change in the scene and cannot update the background image in scenes. The proposed background modeling approach analyzes the illumination change problem. After performing the background subtraction based on the proposed background model, the moving targets can be accurately detected at each frame of the image sequence. In order to produce high accuracy for the motion detection, the binary motion mask can be computed by the proposed threshold function. The experimental analysis based on statistical measurements proves the efficiency of our proposed method in terms of quality and quantity. And it even outperforms substantially existing methods by perceptional evaluation.

Grouping image tokens is an intermediate step needed to arrive at meaningful image representation and summarization. Usually, perceptual cues, for instance, gestalt properties inform token grouping. However, they do not take into account structural continuities that could be derived from other tokens belonging to similar structures irrespective of their location. We propose an image representation that encodes structural constraints emerging from local binary patterns (LBP), which provides a long-distance measure of similarity but in a structurally connected way. Our representation provides a grouping of pixels or larger image tokens that is free of numeric similarity measures and could therefore be extended to nonmetric spaces. The representation lends itself nicely to ubiquitous image processing applications such as connected component labeling and segmentation. We test our proposed representation on the perceptual grouping or segmentation task on the popular Berkeley segmentation dataset (BSD500) that with respect to human segmented images achieves an average F-measure of 0.559. Our algorithm achieves a high average recall of 0.787 and is therefore well-suited to other applications such as object retrieval and category-independent object recognition. The proposed merging heuristic based on levels of singular tree component has shown promising results on the BSD500 dataset and currently ranks 12th among all benchmarked algorithms, but contrary to the others, it requires no data-driven training or specialized preprocessing.

The quest for optimal representations is considered a challenging goal in the field of image processing. This consists of reducing the processing’s complexity while ensuring an efficient reconstruction. An optimal representation should conserve the properties of the image pertaining to smooth content and contours. The multiscale geometric decompositions (MGD) were designed to reach this finality. They were used in many fields and for different purposes, such as feature extraction, detail enhancing, and change detection. A state-of-art of these decompositions is proposed in this paper. We present their theoretical definitions and how they capture the feature of the objects within an image. An overview table is elaborated where we summarize the methods, the data and the different criteria of assessment used in the studied cases. We are interested, particularly, in the use of MGD in a remote sensing (RS) context. Thus, some examples of their applications on RS images are studied. A discussion is presented based on the analyzed cases.

The detection of moving objects in a video sequence is the first step in an automatic video surveillance system. This work proposes an enhancement of a codebook-based algorithm for moving objects extraction. The proposed algorithm used a perceptual-based approach to optimize foreground information extraction complexity by using a modified codebook algorithm. The purpose of the adaptive strategy is to reduce the computational complexity of the foreground detection algorithm while maintaining its global accuracy. In this algorithm, we use a superpixels segmentation approach to model the spatial dependencies between pixels. The processing of the superpixels is controlled to focus it on the superpixels that are near to the possible location of foreground objects. The performance of the proposed algorithm is evaluated and compared to other algorithms of the state of the art using a public dataset that proposes sequences with a dynamic background. Experimental results prove that the proposed algorithm obtained the best the frame processing rate during the foreground detection.

Visual prostheses require an effective representation method due to the limited display condition which has only 2 or 3 levels of grayscale in low resolution. Edges derived from abrupt luminance changes in images carry essential information for object recognition. Typical binary (black and white) edge images have been used to represent features to convey essential information. However, in scenes with a complex cluttered background, the recognition rate of the binary edge images by human observers is limited and additional information is required. The polarity of edges and cusps (black or white features on a gray background) carries important additional information; the polarity may provide shape from shading information missing in the binary edge image. This depth information may be restored by using bipolar edges. We compared object recognition rates from 16 binary edge images and bipolar edge images by 26 subjects to determine the possible impact of bipolar filtering in visual prostheses with 3 or more levels of grayscale. Recognition rates were higher with bipolar edge images and the improvement was significant in scenes with complex backgrounds. The results also suggest that erroneous shape from shading interpretation of bipolar edges resulting from pigment rather than boundaries of shape may confound the recognition.

Existing line segment detectors tend to break up perceptually distinct line segments into multiple segments. We propose an algorithm for merging such broken segments to recover the original perceptually accurate line segments. The algorithm proceeds by grouping line segments on the basis of angular and spatial proximity. Then those line segment pairs within each group that satisfy unique, adaptive mergeability criteria are successively merged to form a single line segment. This process is repeated until no more line segments can be merged. We also propose a method for quantitative comparison of line segment detection algorithms. Results on the York Urban dataset show that our merged line segments are closer to human-marked ground-truth line segments compared to state-of-the-art line segment detection algorithms.

We propose an algorithm for fast computation of three-dimensional (3-D) Hahn moments. First, the symmetry property of Hahn polynomials is provided to decrease the computational complexity at 12%. Second, 3-D Hahn moments are computed by using an algorithm based on matrix multiplication. The proposed algorithm enormously reduces the computational complexity of a 3-D Hahn moment and its inverse moment transform. It can be also implemented easily for high order of moments. The performance of the proposed algorithm is proved through object reconstruction experiments. The experimental results and complexity analysis show that the proposed method outperforms the straightforward method, especially for large size noise-free and noisy 3-D objects.

Visual search is a fundamental technology in the computer vision community. It is difficult to find an object in complex scenes when there exist similar distracters in the background. We propose a target search method in rough three-dimensional-modeling scenes based on a vision salience theory and camera imaging model. We give the definition of salience of objects (or features) and explain the way that salience measurements of objects are calculated. Also, we present one type of search path that guides to the target through salience objects. Along the search path, when the previous objects are localized, the search region of each subsequent object decreases, which is calculated through imaging model and an optimization method. The experimental results indicate that the proposed method is capable of resolving the ambiguities resulting from distracters containing similar visual features with the target, leading to an improvement of search speed by over 50%.

Quantifying image quality without reference is still a challenging problem, especially when different distortions affect the observed image. A no-reference image quality assessment (NR-IQA) metric is proposed. It is based on a fusion scheme of multiple distortion measures. This metric is built in two stages. First, a set of relevant IQA metrics is selected using a particle swarm optimization scheme. Then, a support vector regression (SVR)-based fusion strategy is adopted to derive the overall index of image quality. The obtained results demonstrate clearly that the proposed approach outperforms the state-of-the-art NR-IQA methods. Furthermore, the proposed approach is flexible and could be extended to other distortions.

Imperceptibility and robustness are two key but complementary requirements of any watermarking algorithm. Low-strength watermarking yields high imperceptibility but exhibits poor robustness. High-strength watermarking schemes achieve good robustness but often suffer from embedding distortions resulting in poor visual quality in host media. This paper proposes a unique video watermarking algorithm that offers a fine balance between imperceptibility and robustness using motion compensated wavelet-based visual attention model (VAM). The proposed VAM includes spatial cues for visual saliency as well as temporal cues. The spatial modeling uses the spatial wavelet coefficients while the temporal modeling accounts for both local and global motion to arrive at the spatiotemporal VAM for video. The model is then used to develop a video watermarking algorithm, where a two-level watermarking weighting parameter map is generated from the VAM saliency maps using the saliency model and data are embedded into the host image according to the visual attentiveness of each region. By avoiding higher strength watermarking in the visually attentive region, the resulting watermarked video achieves high perceived visual quality while preserving high robustness. The proposed VAM outperforms the state-of-the-art video visual attention methods in joint saliency detection and low computational complexity performance. For the same embedding distortion, the proposed visual attention-based watermarking achieves up to 39% (nonblind) and 22% (blind) improvement in robustness against H.264/AVC compression, compared to existing watermarking methodology that does not use the VAM. The proposed visual attention-based video watermarking results in visual quality similar to that of low-strength watermarking and a robustness similar to those of high-strength watermarking.

The development of sophisticated sensor technologies gave rise to an interesting variety of data. With the appearance of affordable devices, such as the Microsoft Kinect, depth-maps and three-dimensional data became easily accessible. This attracted many computer vision researchers seeking to exploit this information in classification and recognition tasks. In this work, the problem of face classification in the context of RGB images and depth information (RGB-D images) is addressed. The purpose of this paper is to study and compare some popular techniques for gender recognition and ethnicity classification to understand how much depth data can improve the quality of recognition. Furthermore, we investigate which combination of face descriptors, feature selection methods, and learning techniques is best suited to better exploit RGB-D images. The experimental results show that depth data improve the recognition accuracy for gender and ethnicity classification applications in many use cases.

There are monocular depth cues present in images or videos that aid in depth perception in two-dimensional images or videos. Our objective is to preserve the defocus depth cue present in the videos along with the salient regions during compression application. A method is provided for opportunistic bit allocation during the video compression using visual saliency information comprising both the image features, such as color and contrast, and the defocus-based depth cue. The method is divided into two steps: saliency computation followed by compression. A nonlinear method is used to combine pure and defocus saliency maps to form the final saliency map. Then quantization values are assigned on the basis of these saliency values over a frame. The experimental results show that the proposed scheme yields good results over standard H.264 compression as well as pure and defocus saliency methods.

Textline detection in natural images has been an important problem and researchers have attempted to address this problem by grouping connected components (CCs) into clusters corresponding to textlines. However, developing bottom-up rules that work for multiorientation and/or multiscript textlines is not a simple task. In order to address this problem, we propose a framework that incorporates projection profile analysis (PPA) into the CC-based approach. Specifically, we build a graph of CCs and recursively partition the graph into subgraphs, until textline structures are detected by PPA. Although PPA has been a common technique in document image processing, it was developed for scanned documents, and we also propose a method to compute projection profiles for CCs. Experimental results show that our method is efficient and achieves better or comparable performance on conventional datasets (ICDAR 2011/2013 and MSRA-TD500), and shows promising results on a challenging dataset (ICDAR 2015 incidental text localization dataset).

We present a fully automated approach for smile detection. Faces are detected using a multiview face detector and aligned and scaled using automatically detected eye locations. Then, we use a convolutional neural network (CNN) to determine whether it is a smiling face or not. To this end, we investigate different shallow CNN architectures that can be trained even when the amount of learning data is limited. We evaluate our complete processing pipeline on the largest publicly available image database for smile detection in an uncontrolled scenario. We investigate the robustness of the method to different kinds of geometric transformations (rotation, translation, and scaling) due to imprecise face localization, and to several kinds of distortions (compression, noise, and blur). To the best of our knowledge, this is the first time that this type of investigation has been performed for smile detection. Experimental results show that our proposal outperforms state-of-the-art methods on both high- and low-quality images.

High-speed cameras explore more details than normal cameras in the time sequence, while the conventional video sampling suffers from the trade-off between temporal and spatial resolutions due to the sensor’s physical limitation. Compressive sensing overcomes this obstacle by combining the sampling and compression procedures together. A single-pixel-based real-time video acquisition is proposed to record dynamic scenes, and a fast nonconvex algorithm for the nonconvex sorted ℓ₁ regularization is applied to reconstruct frame differences using few numbers of measurements. Then, an edge-detection-based denoising method is employed to reduce the error in the frame difference image. The experimental results show that the proposed algorithm together with the single-pixel imaging system makes compressive video cameras available.

Due to the particularity of the surface of concrete tunnel lining and the diversity of detection environments such as uneven illumination, smudges, localized rock falls, water leakage, and the inherent seams of the lining structure, existing crack detection algorithms cannot detect real cracks accurately. This paper proposed an algorithm that combines lining seam elimination with the improved percolation detection algorithm based on grid cell analysis for surface crack detection in concrete tunnel lining. First, check the characteristics of pixels within the overlapping grid to remove the background noise and generate the percolation seed map (PSM). Second, cracks are detected based on the PSM by the accelerated percolation algorithm so that the fracture unit areas can be scanned and connected. Finally, the real surface cracks in concrete tunnel lining can be obtained by removing the lining seam and performing percolation denoising. Experimental results show that the proposed algorithm can accurately, quickly, and effectively detect the real surface cracks. Furthermore, it can fill the gap in the existing concrete tunnel lining surface crack detection by removing the lining seam.

A salient object detection method by extensively modeling contextual information in both the saliency feature extraction and the saliency optimization procedure is proposed. First, a sequence of multicontext features is extracted for each segmented image region. This multicontext feature encoding effectively represents the characteristics of image regions and is further mapped to the initial saliency value estimation using a nonlinear regressor. Second, contextual information is also utilized to optimize the initial saliency map, which is realized by constructing a region-level conditional random field (CRF). As such, the quality of the initial coarse saliency maps is promoted in a more principled manner. Third, multiple CRFs, defined over different scales of segmentation, are calculated and integrated so that different ranges of contextual information could contribute to the saliency optimization. Eventually, consistent saliency maps with uniformly highlighted salient regions and clear boundaries are generated. The proposed method is extensively evaluated on three public benchmark datasets, and experimental results demonstrate that our method can produce promising performance when compared to state-of-the-art salient object detection approaches.

Bandwidth saving, power consumption, and fill factor improvement are known as vitally important challenges image sensor designers face in order to accomplish high-performance imaging systems. This paper presents an ultralow power, high fill factor smart complementary metal oxide semiconductor (CMOS) image sensor with motion detection capability. In this efficient methodology, the amount of redundant data processed in unimportant frames has been reduced significantly, and therefore, the proposed imaging system consumes less power compared with counterpart imagers. Furthermore, a pixel structure is introduced that outputs two consecutive frame voltages in series, with the result that the pixel size is minimized and a higher fill factor is achieved. In order to simulate the image capturing procedure, a state-of-the-art approach based on MATLAB and HSPICE software is devised, which is another important achievement of this paper. The performance of this technique is demonstrated using a 64×64  pixel sensor designed in a 0.18-μm standard CMOS technology. The sensor chip consumes 0.2 mW of power while operating at 100 fps with a fill factor of 45%.

An efficient biprediction decision scheme of high efficiency video coding (HEVC) is proposed for fast-encoding applications. For low-delay video applications, bidirectional prediction can be used to increase compression performance efficiently with previous reference frames. However, at the same time, the computational complexity of the HEVC encoder is significantly increased due to the additional biprediction search. Although a some research has attempted to reduce this complexity, whether the prediction is strongly related to both motion complexity and prediction modes in a coding unit has not yet been investigated. A method that avoids most compression-inefficient search points is proposed so that the computational complexity of the motion estimation process can be dramatically decreased. To determine if biprediction is critical, the proposed method exploits the stochastic correlation of the context of prediction units (PUs): the direction of a PU and the accuracy of a motion vector. Through experimental results, the proposed method showed that the time complexity of biprediction can be reduced to 30% on average, outperforming existing methods in view of encoding time, number of function calls, and memory access.

Feature correspondence is one of the essential difficulties in image processing, given that it is applied within a wide range in computer vision. Even though it has been studied for many years, feature correspondence is still far from being ideal. This paper proposes a multigeometric-constraint algorithm for finding correspondences between two sets of features. It does so by considering interior angles and edge lengths of triangles formed by third-order tuples of points. Multigeometric-constraints are formulated using matrices representing triangle similarities. The experimental evaluation showed that the multigeometric-constraint algorithm can significantly improve the matching precision and is robust to most geometric and photometric transformations including rotation, scale change, blur, viewpoint change, and JPEG compression as well as illumination change. The multigeometric-constraint algorithm was applied to object recognition which includes extraprocessing and affine transformation. The results showed that this approach works well for this recognition.

Various methods are proposed for license plate recognition, but none of them are universal. Some common methods for license plate localization, character extraction, and recognition are analyzed. Then a system is proposed to recognize the Bahamian license plate with touching characters. A vertical edge-based method with a modified sliding window technique is used to locate the license plate, and a machine learning process is used to trim the region. The located license plate is rectified by using the minimum enclosing box and the stroke width value. Then the vertical projection and pairs of extreme points are combined to segment the characters. Finally, a deep learning method is used to recognize the characters. 2996 images are experimented on and the total recognition accuracy achieves 83.29%. Typical methods of each stage are implemented to compare with the proposed methods. In addition, the proposed system is experimented on a public dataset to show the generalization ability of the system. The experimental results show that the proposed system performs better than the other methods and is able to be used in a real-time application.

Video sequences captured over a long range through the turbulent atmosphere contain some degree of atmospheric turbulence degradation (ATD). Stabilization of the geometric distortions present in video sequences containing ATD and containing objects undergoing real motion is a challenging task. This is due to the difficulty of discriminating which part of visible motion is real motion and which part is caused by ATD warping. Due to this, most stabilization techniques applied to ATD sequences distort real motion in the sequence. We propose a method to classify foreground regions in ATD video sequences. This classification is used to stabilize the background of the scene while preserving objects undergoing real motion by compositing them back into the sequence. A hand-annotated dataset of three ATD sequences is produced with which the performance of this approach can be quantitatively measured and compared against the current state of the art.

Depth modeling mode 1 (DMM-1) processes its prediction unit (PU) without taking into account its neighborhood. However, an extensive analysis of neighbor pattern selection showed that the processing PU could use the same pattern of its neighbors or an adaptation of these patterns. Therefore, this work proposes the DMM-1 fast pattern selector (DFPS) algorithm that includes lightweight and medium-weight DMM-1 pattern predictors. DFPS starts using the lightweight predictor, whose output is compared with a threshold and then the algorithm employs a DMM-1 refinement or the medium-weight predictor. The result of this last predictor is compared against a new threshold, where the encoder decides if the remaining patterns of the DMM-1 algorithm or only its refinement should be evaluated. Experiments evaluating DFPS encoding efficiency in the all-intra mode demonstrate that by reducing 71% and 84% of the DMM-1 patterns’ evaluation, the complexity is reduced 11.7% and 13.4%, respectively, without significantly affecting the quality of the synthesized views.

We present a fast and minimum delay algorithm for detecting the pupil center, called the “cross spread” tracking technique. This algorithm is meant for video eye trackers that estimate gaze direction from the position of the pupil center in the captured images. Contrary to other solutions, we do not try to make this technique robust to distractors such as reflections, distortions caused by glasses, or eyelids covering the pupil, but rather we assume eye tracking in stable light conditions. We argue that this approach is useful in many eye tracking applications, such as gaze tracking during psychophysical experiments in stable laboratory conditions, and that this approach can significantly reduce the eye tracker’s complexity while maintaining its accuracy and performance. The proposed cross spread technique estimates pupil by tracing rays in horizontal and vertical directions in the image, starting from a point in the pupil region and continuing to the pupil boundary. The found boundary points determine the next starting point and the procedure is iteratively repeated. Parallel processing can be efficiently used enabling accurate pupil center detection in <2  ms on typical laptops. We compare the proposed algorithm to other pupil detection algorithms.

The sparsifying representation plays a significant role in compressive sensing (CS)-based hyperspectral (HS) imaging. Training the dictionaries for each dimension from HS samples is very beneficial to accurate reconstruction. However, the tensor dictionary learning algorithms are limited by a great amount of computation and convergence difficulties. We propose a least squares (LS) type multidimensional dictionary learning algorithm for CS-based HS imaging. We develop a practical method for the dictionary updating stage, which avoids the use of the Kronecker product and thus has lower computation complexity. To guarantee the convergence, we add a pruning stage to the algorithm to ensure the similarity and relativity among data in the spectral dimension. Our experimental results demonstrated that the dictionaries trained using the proposed algorithm performed better at CS-based HS image reconstruction than those trained with traditional LS-type dictionary learning algorithms and the commonly used analytical dictionaries.

The projective model is an important mapping function for the calculation of global transformation between two images. However, its hardware implementation is challenging because of a large number of coefficients with different required precisions for fixed point representation. A VLSI hardware architecture is proposed for the calculation of a global projective model between input and reference images and refining false matches using random sample consensus (RANSAC) algorithm. To make the hardware implementation feasible, it is proved that the calculation of the projective model can be divided into four submodels comprising two translations, an affine model and a simpler projective mapping. This approach makes the hardware implementation feasible and considerably reduces the required number of bits for fixed point representation of model coefficients and intermediate variables. The proposed hardware architecture for the calculation of a global projective model using the RANSAC algorithm was implemented using Verilog hardware description language and the functionality of the design was validated through several experiments. The proposed architecture was synthesized by using an application-specific integrated circuit digital design flow utilizing 180-nm CMOS technology as well as a Virtex-6 field programmable gate array. Experimental results confirm the efficiency of the proposed hardware architecture in comparison with software implementation.

The visible quality assessment of images is important to evaluate the performance of image processing methods such as image correction, compressing, and enhancement. The structural similarity is widely used to determine the visible quality; however, existing structural similarity metrics cannot correctly assess the perceived human visibility of images that have been slightly geometrically transformed or images that have undergone significant regional distortion. We propose an improved structural similarity metric that is more close to human visible evaluation. Compared with the existing metrics, the proposed method can more correctly evaluate the similarity between an original image and various distorted images.

The research in developing steganalytic features has been highly successful. These features are extremely powerful when applied to supervised binary classification problems. However, they are incompatible with unsupervised universal steganalysis because the unsupervised method cannot distinguish embedding distortion from varying levels of noises caused by cover variation. This study attempts to alleviate the problem by introducing similarity retrieval of image statistical properties (SRISP), with the specific aim of mitigating the effect of cover variation on the existing steganalytic features. First, cover images with some statistical properties similar to those of a given test image are searched from a retrieval cover database to establish an aided sample set. Then, unsupervised outlier detection is performed on a test set composed of the given test image and its aided sample set to determine the type (cover or stego) of the given test image. Our proposed framework, called SRISP-aided unsupervised outlier detection, requires no training. Thus, it does not suffer from model mismatch mess. Compared with prior unsupervised outlier detectors that do not consider SRISP, the proposed framework not only retains the universality but also exhibits superior performance when applied to high-dimensional steganalytic features.

As a typical multilinear dimensionality reduction (DR) method, tensor locality preserving projection (TLPP) has been successfully applied in many practical problems. However, TLPP depends mainly on preserving its local neighbor graph which often suffers from the following issues: (1) the neighbor graph is constructed with the Euclidean distance which fails to consider the relationships among different coordinates for tensor data; (2) the affinity matrix only focuses on the local structure information of samples while ignoring the existing label information; (3) the projection matrices are nonorthogonal, thus it is difficult to preserve the local manifold structure. To address these problems, a multilinear DR algorithm called optimal neighbor graph-based orthogonal tensor locality preserving projection (OG-OTLPP) is proposed. In OG-OTLPP, an optimal neighbor graph is first built according to tensor distance. Then the affinity matrix of data is defined by utilizing both the label information and the intrinsic local geometric properties of the data. Finally, in order to improve the manifold preserving ability, an efficient and stable scheme is designed to iteratively learn the orthogonal projections. We evaluate the proposed algorithm by applying it to image recognition. The experimental results on five public image databases demonstrate the effectiveness of our algorithm.

This paper presents a scheme of principal node analysis (PNA) with the aim to improve the representativeness of the learned codebook so as to enhance the classification rate of scene image. Original images are normalized into gray ones and the scale-invariant feature transform (SIFT) descriptors are extracted from each image in the preprocessing stage. Then, the PNA-based scheme is applied to the SIFT descriptors with iteration and selection algorithms. The principal nodes of each image are selected through spatial analysis of the SIFT descriptors with Manhattan distance (L1 norm) and Euclidean distance (L2 norm) in order to increase the representativeness of the codebook. With the purpose of evaluating the performance of our scheme, the feature vector of the image is calculated by two baseline methods after the codebook is constructed. The L1-PNA- and L2-PNA-based baseline methods are tested and compared with different scales of codebooks over three public scene image databases. The experimental results show the effectiveness of the proposed scheme of PNA with a higher categorization rate.

We focus on the problem of pose-based gait recognition. Our contribution is two-fold. First, we incorporate a local histogram descriptor that allows us to encode the trajectories of selected limbs via a one-dimensional version of histogram of oriented gradients features. In this way, a gait sequence is encoded into a sequence of local gradient descriptors. Second, we utilize a robust encoding method in which the residuals of local descriptors, with respect to a discriminative model, are aggregated into fixed length vectors. This technique combines the advantages of both residual aggregation and soft-assignment techniques, resulting in a powerful vector representation. For classification purposes, we use a nonlinear kernel to map vectors into a reproducing kernel Hilbert space. Then, we classify an encoded gait sequence according to the sparse representation-based classification method. Experimental evaluation on two publicly available datasets demonstrates the effectiveness of the proposed scheme on both recognition and verification tasks.

Hyperspectral cameras provide additional information in terms of multiple sampling of the visible spectrum, holding information that could be potentially useful for biometric applications. This paper investigates whether the performance of hyperspectral face recognition algorithms can be improved by considering single and multiple one-dimensional (1-D) projections of the whole spectral data along the spectral dimension. Three different projections are investigated and found by optimization: single-spectral band selection, nonnegative spectral band combination, and unbounded spectral band combination. Since 1-D projections can be performed directly on the imaging device with color filters, projections are also restricted to be physically plausible. The experiments are performed on a standard hyperspectral dataset and the obtained results outperform eight existing hyperspectral face recognition algorithms.

A fast measurement matrix based on scrambled block Hadamard ensemble for compressed sensing (CS) of hyperspectral images (HSI) is investigated. The proposed measurement matrix offers several attractive features. First, the proposed measurement matrix possesses Gaussian behavior, which illustrates that the matrix is universal and requires a near-optimal number of samples for exact reconstruction. In addition, it could be easily implemented in the optical domain due to its integer-valued elements. More importantly, the measurement matrix only needs small memory for storage in the sampling process. Experimental results on HSIs reveal that the reconstruction performance of the proposed measurement matrix is comparable or better than Gaussian matrix and Bernoulli matrix using different reconstruction algorithms while consuming less computational time. The proposed matrix could be used in CS of HSI, which would save the storage memory on board, improve the sampling efficiency, and ameliorate the reconstruction quality.

The advance of video coding technology enables multiview video (MVV) or three-dimensional television (3-D TV) display for users with or without glasses. For mobile devices or wireless applications, a distributed video coder (DVC) can be utilized to shift the encoder complexity to decoder under the MVV coding framework, denoted as multiview distributed video coding (MDVC). We proposed to exploit both inter- and intraview video correlations to enhance side information (SI) and improve the MDVC performance: (1) based on the multiview motion estimation (MVME) framework, a categorized block matching prediction with fidelity weights (COMPETE) was proposed to yield a high quality SI frame for better DVC reconstructed images. (2) The block transform coefficient properties, i.e., DCs and ACs, were exploited to design the priority rate control for the turbo code, such that the DVC decoding can be carried out with fewest parity bits. In comparison, the proposed COMPETE method demonstrated lower time complexity, while presenting better reconstructed video quality. Simulations show that the proposed COMPETE can reduce the time complexity of MVME to 1.29 to 2.56 times smaller, as compared to previous hybrid MVME methods, while the image peak signal to noise ratios (PSNRs) of a decoded video can be improved 0.2 to 3.5 dB, as compared to H.264/AVC intracoding.

Compressive imaging presents the challenge of reconstructing an image accurately from its under-sampling measurements. The total variation model, used as the regularization for imaging reconstruction, ignores structure direction, which tends to result in over-smoothing of homogenous regions and staircase artifacts in textured regions. This paper proposes to solve this problem using a compressive reconstruction method with adaptive directional total variation (DTV) based on the structure orientation field (SOF). This approach builds the estimation model of the structure tensor upon the Huber norm as the penalty term, reducing the degradation influence and reaching a quick solution using the primal-dual algorithm. The method then uses the structure tensor to describe the SOF and to establish compressive reconstruction in which the SOF restricts the DTV regularization. Finally, by analyzing the structure of the objective function, this approach adopts the linearized alternating direction method of multipliers to address the challenges presented by the reconstruction model. The experimental results show that the proposed method requires fewer iterations and demonstrates better reconstruction performance than existing methods, as indicated by examining texture details for visual fidelity and quality as measured through other objective criteria.

Support vector machine (SVM) classifiers are widely applied to hyperspectral image (HSI) classification and provide significant advantages in terms of accuracy, simplicity, and robustness. SVM is a well-known learning algorithm that maximizes the minimum margin. However, recent theoretical results pointed out that maximizing the minimum margin leads to a lower generalization performance than optimizing the margin distribution, and proved that the margin distribution is more important. In this paper, a large margin distribution machine (LDM) is applied to HSI classification, and optimizing the margin distribution achieves a better generalization performance than SVM. Since the raw HSI feature space is not the most effective space for representing HSI, we adopt factor analysis to learn an effective HSI feature and the learned features are further filtered by a structure-preserved filter to fully exploit the spatial structure information of HSI. The spatial structure information is integrated in the feature learning process to obtain a better HSI feature. Then we propose a multiclass LDM to classify the filtered HSI feature. Experimental results show that the proposed LDM with feature learning method achieves the classification performance of the state-of-the-art methods in terms of visual quality and three quantitative evaluations and indicates that LDM has a high generalization performance.

We present a high-performance improvement method for implementation of local processing algorithms for video frames by using the benefit of memoization technique. Memoization is a technique that uses the advantage of data redundancy to minimize the amount of computations performed by retrieving previous results instead of computing again, which leads to faster processing speed. In this method, the benefit of interframe redundancy in adjacent frames is used for memoizing where the pixels in a sequence of frames are correlated. We have developed this method in software and applied it to edge detection and median filters. The typical speedups achieved in the median filter range from 2.2× with exact results to 6.48× in tolerant methods and in edge detection filter range from 2.1× with exact results to 5.41× in tolerant method. The structural similarity index metric that is used for evaluating the perceived similarity of the tolerant result with ideal result was applied to each adjacent frame in sample stream frames. The typical values of this parameter were 0.96 in median filter and 0.71 in edge detection filter.

This PDF file contains the errata for “JEI Vol. 25 Issue 06 Paper JEI-2016-1118-ERR” for JEI Vol. 25 Issue 06