Revolutionising Image Enhancement Leveraging Power OF CNN’S

Image restoration has several uses, remote sensing, surveillance, medical imaging, and computational photography. Recovering high-quality image content from lower-quality images is its goal. Convolutional neural networks (CNN’s) have shown better results than the existing methods for image restoration in amazing ways. Many of the present CNN-based methods work with either fully-resolution or lower-resolution representations. While the second scenario produces acceptably trustworthy but spatially less accurate outputs, the first instance gives spatially the same but contextually less accurate results, Image restoration has several uses, in the medical field, traffic policies and criminal investigation, and astronomical image enhancement Its purpose is to recover greater quality image content from its old version, which has less quality So, having the overall aim of maintaining spatially intact higher quality representations over the whole network and gaining the maximum contextual information from the previous lower resolution representations. In previous projects, they used RNN and CNN to get the output we are using CNN and aim for better output by doing the project in our kind of way. Our project’s crucial component is a multi-scale residual block, which is made up of several vital components, including (a)exchange of information between the multiresolution streams, (b) channel and special attention mechanisms for contextual information capturing (c) parallel multi-resolution convolution streams for the extracting multi-scaled features,(d) attention multi-scale feature aggregation. In brief, our project will aim towards learning a bunch of features that will maintain higher resolution spatial information and details while combining contextual data from various scales. Thereby increasing the image quality to a greater version resolution, physical constraints, low dynamic range, downsampling, spatial features, and receptive area.