With the development of deep learning, deep neural network can be consumed to many kinds of tasks, such as target detection, super resolution and other image processing applications. However, super resolution for sonar images adopts traditional ways without neural network. Due to the characteristic of sonar sensors, which shoot images in turbid underwater environment and have low resolution, objects taken by sonar sensors are difficult to recognize. This paper proposes an efficient method called Multi-stage Residual Network (MRN) which combines neural network to achieve super resolution for sonar images. Adding pixel shuffle to the medium of the structure, the number of layers and blocks in every part is different. In addition, we trained the network with no modification from underwater sonar images. Experimental results show that low-resolution fuzzy images can acquire a clear super resolution by using our model, and furthermore, PSNR of our resultant images is higher than that of the interpolation algorithms and residual network.
Facial illumination severely affects the face recognition performance; thus, it should be finely compensated beforehand. This paper mainly studies three representative illumination-insensitive representation methods including GRF, WF and LBP, all of which consider using the local pattern information to extract facial features. We first present the ideas and theories of each method. Then based on the reflectance model, the underlying connections of GRF and WF are discussed through showing the deduction of GRF and WF how they exclude the luminance component and are only related to the intrinsic facial features. We also give the explanation about the correspondence of LBP to the reflectance model. Finally, experiments on a standard but challenging illuminated face database are conducted, in which GRF, WF and LBP are tested and compared to other illumination normalization methods in terms of face recognition rate.
Compensating the illumination of a face image is an important process to achieve effective face recognition under severe illumination conditions. This paper present a novel illumination normalization method which specifically considers removing the illumination boundaries as well as reducing the regional illumination. We begin with the analysis of the commonly used reflectance model and then expatiate the hybrid usage of adaptive non-local smoothing and the local information coding based on Weber’s law. The effectiveness and advantages of this combination are evidenced visually and experimentally. Results on Extended YaleB database show its better performance than several other famous methods.
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.