K-Multi Constraint Shortest Paths for Video Compression in Handheld Devices Using Improved Quadtree based Motion Estimation

Abstract

Today, video compression is one of promising field which finds its applications in many areas such as mobile video, videophones, Digital TV broadcasting, surveillance system etc. However, transmission of video over handheld devices is increasing in many folds in the recent time. The biggest challenge for transmission of video over these devices is their limited battery life as well as restricted bandwidth. The biggest advantage of handheld devices is that compromise of picture quality is acceptable to ease the bandwidth requirement. So, video coding standards which allow degradation in quality of reconstructed video may be used so that low-bandwidth requirement is fulfilled. One of main component of video coding standards, responsible for reducing bandwidth requirement, is motion estimation (ME) which consume almost 80-90% of total power of device. In this paper, an improved quadtree based ME algorithm is proposed to meet the requirements of handheld devices. Pre-processing of video frame based on the threshold of edge intensity is applied to motion estimation so that active and dead blocks are calculated based on some threshold value of edge intensity. After finding motion vector, Lagrange optimization is used to find optimal solution for rate distortion performance. K-multi constraint shortest paths are calculated using AI search algorithms. Experimental results show that different paths require different value of displaced vector field (DVF) and displacement field difference (DFD) providing flexibility to end user to select the paths according their bit budget. Simulation is done for different threshold values at different frame rate and it shows that bit can be saved using improved quadtree based ME algorithm.