Efficient Video Compression Schemes

Abstract

The wide use of images and videos in the day to day communication in recent trends made video compression a significant feature in information transmission and social networking. Also the limited bandwidth for transmission and limited memory make it a serious phenomenon to consider. There is a need to improve the video encoding process which can encode the video data with low computational complexity and high compression ratio along with maintaining quality. Motion Estimation (ME) is the widely used scheme in many encoders like MPEG-2, MPEG-4 and H.264 that removes temporal redundancy. In this thesis, many ME algorithms like Full Search, Logarithmic Search, Diamond Search, Cross Diamond Search, Kite Cross Diamond Search (KCDS), Hexagonal Search (HEXS), Enhanced Hexagonal Search (ENHEXS) etc. are implemented and analysed. Based on the conclusions drawn modifications to KCDS and HEXS are proposed by imparting the concept of motion vector prediction. A novel Hybrid Hexagonal Kite Cross Diamond Search (HYBHKS) algorithm is proposed. It has the capability of adaptive switching of search patterns based on the type of motion of the block that can reduce the computational complexity of the encoder maintaining quality. Analysis is done on the impartation of Discrete Hartley Transform in video encoding. Video Compression may still be increased by reducing the information to be encoded. This is done by skipping some irrelevant information. In this thesis, compression techniques involving the concepts of spatial and temporal correlation of rows are implemented where only alternate rows of the video are fed to the video encoder. At the receiver the decoded video is resized to original dimensions by predicting the skipped rows. Also analysis is also done on down sampling and up-sampling of video before and after encoder and decoder respectively. Unsharp masking is implemented and a new technique for edge boosting using DWT is proposed.