Article
Image/Video Super Resolution Using CNN And Auto encoders
For the purpose of improving the coding efficiency of lossy compressed films, this study presents a groundbreaking deep learning approach. More specifically, it makes use of a Variable-Filter-Size Residue-Learning Convolutional Neural Network with Encoders. Through the utilization of this cuttingedge method, common distortions and artifacts like as blocking, blurring, and ringing are addressed. By surpassing the limitations of the High Efficiency Video Coding (HEVC) standard, our model is able to make significant improvements in both the quality of the video and the efficiency with which it compresses the video. In addition, we present a novel approach to the problem of picture super-resolution by making use of Convolutional Neural Networks (CNNs) and auto encoders. A deep convolutional neural network (CNN) model is trained in conjunction with an auto encoder architecture through the utilization of paired datasets consisting of high-resolution and low-resolution pictures using our methods. Throughout the training process, the CNN is able to extract high-level features from low-resolution photos, while the auto encoder is able to learn how to properly rebuild high-resolution images, therefore capturing delicate details and textures. During the inference phase, our trained model takes an input image with a low resolution and creates an output image with a high resolution that corresponds to the same input image.
Full Text Attachment
