Optimized Convolutional Neural Network for High Quality Image Super Resolution

Abstract---- - One of the most crucial methods for digital image processing is super-resolution. In this method, one sub category is Single Image Super Resolution (SISR) and other is Multi-Frame Super Resolution (MFSR). A low-resolution image shall be used by SISR to output a high-resolution image whereas in MFSR various images of the same view with slight variations in positions are considered collectively as input to produce a single high- resolution image. In our paper Peak Signal Noise Ratio (PSNR), Structural Similarity Index Metric (SSIM), Multi Scale Structural Similarity Index Metric (MSSSIM) and Weighted Peak Signal Noise Ratio (WPSNR) are the objective metrics used to evaluate Very Deep Super Resolution Network (VDSR) and Super Resolution Convolutional Neural Network (SRCNN) models of image super resolution Keywords—Image Super-Resolution, Convolutional Neural Network, Lightweight Networks, Deep Learning, LESRCNN, Structural Similarity Index Metric, Peak Signal Noise Ratio, Multi Scale Structural Similarity Index Metric and Weighted Peak Signal Noise Ratio