Publication

Abstract : Blind Image Super-Resolution (BISR) is a widely recognized image processing the approach. BISR, which seeks to super-resolving low-resolution images with undetermined deterioration, is gaining prominence due to its usefulness in stimulating real-world applications. In this research, we developed the Wavelet-Subpixel Residual Attention Network (WSRAN) for BISR, which significantly increases SR efficiency without sacrificing performance while enhancing generalized ability to real-world damaged pictures. The subpixel convolution layer converts LR feature maps into reconstructed outputs. We also employ the Residual Attention Block (RAB). Extensive trials indicate that when compared to SR approaches, our WSRAN requires fewer parameters and achieves statistically and subjectively competitive results. In the future, we want to find a means to increase the upscale factor beyond three.