A DEEP LEARNING MODELVIDEO COMPRESSION FOR HIGH EFFICIENCY VIDEO CODING
Abstract
The evolution of the internet and related communication technologies is having a significant impact on video conferencing and compressing techniques. According to these standards, motion estimation and compression are required to remove the lag time between successive frames in video sequences. The high efficiency video coding (HEVC) plays the major role in these applications, where advanced versions of HEVC provided better outcomes. This indicates that the approach can handle a wide variety of search patterns, from very tiny to very big, depending on the total number of check sites. Starting from the middle of a content block, the deep learning-based content block test zone search (DL-CBTZS) technique performs a diamond-shaped split. The origin of certain video frames may prevent them from having sufficient assessment criteria. If the motion vector acquired by the DL-CBTZS results in a lesser error, then the DL-CBTZS will converge to block searches after content identification. Here, the deep learning model is used for inter and intra frame prediction. As a result, it's a time-consuming task. DL-CBTZS is a fast and effective technique for generating sensible search patterns. Using DL-CBTZS and the Hybrid Wavelet Transform method, a parallel transformation strategy was created. The development of this technique was motivated by the need to lessen the footprint of video files. Simulated results corroborate the superiority of the suggested method over the conventional methods.
Keywords
High efficiency video coding, deep learning, content block test zone search, inter frame prediction, intra frame prediction.