Super Resolution Reconstruction of Low Resolution Video Using Sparse Technique
Rohita H. Jagdale, Sanjeevani Kiran Shah · 2019
Super resolution (SR) deals with enhancing the quality of low resolution (LR) image to obtain high resolution (HR) image. It is required to design the system for SR using sparse representation method. Sparsity-based SR techniques deal with spacing the pixels of LR image and filling them by estimating new pixels. SR based on sparsity concept, collective with structure modulated technique, upgrades its modeling capability. The technique can be beneficial to favor the pursuance of image SR based on sparsity. Patch wise each LR frame is divided and processed iteratively for better results. Firstly, LR frame of the video input is given to the system. After converting it into YCbCr color space with bi-cubic interpolation; LR patch features are extracted and corresponding HR patches will be generated. After iterative back-projection, HR image is converted back to RGB color space and given out as the HR sparse output. It is ambitious to the state-of-the-art methods like Bicubic, SRCNN, ASDS, NCSR and NA-VSR, both visually and quantitatively. The quality of the HR image is decided by PSNR and SSIM values.