When it comes to image classification at a device terminal a traditional machine learning method tends to pose a risk of privacy disclosure, while the federated learning is able to alleviate such privacy problem to a certain extent by saving device data locally to train the local model. In this paper images are classified based on horizontal federated learning and FedAvg optimization algorithm. Classification model train is carried out based upon CIFAR-10 dataset and ResNet-18. When the learning rate is appropriate, the optimization algorithm adopted has faster convergence, fewer communication rounds and better classification effect. The algorithm remains convergent and even shows faster convergence in the case of heterogeneous data.
X-ray angiograms, which suffer from low-contrast and noise, need to be improved by both the image enhancement and denoising techniques. However, the goals of these two tasks usually conflict, which makes it difficult to efficiently combine the enhancement and denoising in one scheme. To solve this problem, we propose a novel spatial-frequency filtering (SFF) scheme to simultaneously enhance and denoise low-quality X-ray cardiovascular angiogram images. The proposed scheme includes three key components: Firstly, a relative total variation method is employed as a guide filter to separate an input image into two parts, including the base layer with strong structures and the detail layer with weak structures and noise. Then the base layer is enhanced by a proposed improved histogram equalization (IHE) method while the detail layer is extracted by a short-time Fourier transform and is further enhanced by using a proposed adaptive correction parameter. Finally, the improved image is the combination of results obtained by the two components. Both quantitative and qualitative results of experiments on real-world low-quality X-ray angiogram images demonstrate that the proposed method outperforms the state-of-the-arts in terms of contrast enhancement, structure preservation, and noise reduction.
Stripe is a common degradation phenomenon in remote sensing images. The variation-based de-striping method, due to the defect of the model itself, always has an unnecessary influence on the stripe-free area while correcting the stripe, and cannot satisfy some requirements in high-precision quantitative applications or sensitive data processing of remote sensing images. This paper proposes a high-precision stripe correction method, which first detects the position of the stripes, and then uses the interpolation idea to correct the stripe to solve the fidelity problem of the stripe-free area in the de-striping process. We use the rational assumption that the derivative of the real signal in the stripe region (to be repaired) is consistent with the derivative of the observed signal, and then selects cubic Hermite spline interpolation method for de-striping, which can uses the derivative information of the region to be repaired (ie, the derivative information of the stripe region) to overcoming the difficulty of the existing interpolation de-stripe method not being able to work well when the stripes is too wide. The experimental results show that our method can effectively remove the stripes and maintain the stripe-free area intact.
Denoising algorithms based on gradient dependent energy functionals, such as Perona-Malik, total variation and adaptive total variation denoising, modify images towards piecewise constant functions. Although edge sharpness and location is well preserved, important information, encoded in image features like textures or certain details, is often compromised in the process of denoising. In this paper, We propose a novel Spatially Adaptive Guide-Filtering Total Variation (SAGFTV) regularization with image restoration algorithm for denoising images. The guide-filter is extended to the variational formulations of imaging problem, and the spatially adaptive operator can easily distinguish flat areas from texture areas. Our simulating experiments show the improvement of peak signal noise ratio (PSNR), root mean square error (RMSE) and structure similarity increment measurement (SSIM) over other prior algorithms. The results of both simulating and practical experiments are more appealing visually. This type of processing can be used for a variety of tasks in PDE-based image processing and computer vision, and is stable and meaningful from a mathematical viewpoint.
Total variation(TV) based on regularization has been proven as a popular and effective model for image restoration, because of its ability of edge preserved. However, as the TV favors a piece-wise constant solution, the processing results in the flat regions of the image are easily produced "staircase effects", and the amplitude of the edges will be underestimated; the underlying cause of the problem is that the regularization parameter can not be changeable with spatial local information of image. In this paper, we propose a novel Scatter-matrix eigenvalues-based TV(SMETV) regularization with image blind restoration algorithm for deblurring medical images. The spatial information in different image regions is incorporated into regularization by using the edge indicator called difference eigenvalue to distinguish edges from flat areas. The proposed algorithm can effectively reduce the noise in flat regions as well as preserve the edge and detailed information. Moreover, it becomes more robust with the change of the regularization parameter. Extensive experiments demonstrate that the proposed approach produces results superior to most methods in both visual image quality and quantitative measures.
In the application of image processing and pattern recognition, the precision of image preprocessing has a great influence on the image after-processing and analysis. This paper describes a novel local double mean weighted algorithm (hereinafter referred to as D-M algorithm) for image denoising. Firstly, the pixel difference and the absolute value are taken for the current pixels and the pixels in the neighborhood; then the absolute values are sorted again, the means of such pixels are taken in an half-to-half way; finally the weighting coefficient of the mean is taken. According to a large number of experiments, such algorithm not only introduces a certain robustness, but also improves increment significantly.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.