Research on Video Compression Algorithm Based on Deep Learning

Abstract

The storage of any type of data on the web requires the use of that particular type of data. The amount of pictures, movies, and other forms of content that are similar to one another that are circulating on the internet has skyrocketed. Even under the weight of many resource constraints, such as bandwidth bottlenecks and noisy channels, users of the Internet demand the data they access to be easily understood. As a direct result of this, data compression is rapidly becoming an increasingly important topic among the larger engineering community. Using deep neural networks for the purpose of data compression has been the subject of some previous research. Several different machine learning approaches are now being implemented into data compression strategies and put to the test in an effort to achieve improved lossy and lossless compression outcomes. The typical video compressive sensing reconstruction algorithm has an excessively lengthy delay throughout the reconstruction process. It is unable to make full use of the spatial and temporal correlation of video, which results in an improvement in the quality of the reconstruction poor. In this research, a video compression method that is based on deep learning (DL) is proposed, and it has the potential to handle these challenges effectively.