PROJECT TITLE :

Soft-Edge Assisted Network for Single Image Super-Resolution

ABSTRACT:

In order to solve the problem of single image super-resolution (SISR), it is necessary to deal with a difficult inverse problem. Previous CNN-based SR approaches tend to directly train the mapping from low-resolution image to high resolution image through some complicated convolutional Neural Networks. However, increasing the network's depth haphazardly is not the optimal approach because the gains in performance are minimal and the computing costs are enormous. To aid with picture reconstruction, it is preferable to use the model's prior knowledge of the image as input. Many computer vision tasks have relied on the soft-edge as a key image component. Soft-edge aided Network (SeaNet) is proposed in this paper to reconstruct high-quality SR images using image soft-edge. A rough image reconstruction network (RIRN), a soft-edge reconstruction network (Edge-Net), and an image refinement network are all part of the SeaNet suggested by researchers (IRN). A two-stage reconstruction is required to finish the project. It is at this point in Stage-I that the RIRN and the Edge-Net are used to create approximate feature maps of the SR soft-edge. All preceding stages' outputs are combined in Stage II and sent to the IRN for high-quality SR reconstruction. The use of picture soft-edge aids our SeaNet in fast converging and performing at its best, as demonstrated in numerous studies.