CONVOLUTIONAL SPARSE CODING FOR NONLOCAL IMAGE RESTORATION AND SUPER RESOLUTION

Convolutional sparse coding and Simultaneous sparse coding are gives great potential in various low level tasks. It leads to several state of art restoration technique in image processing. convolution sparse coding (CSC) a both the approach have shown great potential in various low-level. In convolutional sparse coding involves 3 type parameter must learn : (1) Decompose the low resolution image into low resolution (LR) sparse feature maps using a set of filters, (2). Predict the mapping function from LR to High resolution (HR) feature maps and (3) Reconstruct from the predicted HR feature maps to the HR images using set of filters via simple convolution operations. In different estimation perspective, like decomposition of similar packed patches can be viewed as pooling both non-local and local information for estimating signal variances. So, it inspires me to develop a new class of image restoration algorithms based on CSV approach. Our R&D work results compare favorably with those obtained by existing techniques, especially with a large amount of missing data and at high noise levels