Research On Video Stabilization Algorithm Based On Coding Theory

Video coding aims to reduce the redundancy of the raw video data and video stabilization targets at removing shaky camera motions in the video signal.In this work,we focus on the design of high-efficiency video coding and video stabilization to achieve a more effective processing on the video data obtained from the mobile video capture devices.In this work,we firstly focus on the design of a new compression method for the color image and video signals.To achieve this goal,we build up a cross-space distortion directed downsampling(CSDDD)to implement an efficient color image compression.This proposed downsampling scheme is built upon the RGB-to-YCbCr conversion and used to generate the 4:2:0 YCbCr image signal.When it is applied to the color image compression,an significant quality gain is achieved.Then,we propose a new method for video stabilization based on the video coding.In the traditional video stabilization,the camera motions are often estimated based on the image feature tracking.However,the feature trackingis a time-consuming job and is prone to errors.Meanwhile,it is noticed that the captured video signalsmustbe compressedbefore the further processing.Therefore,we may use the motion vectors of the video coding to estimate the camera motions.Compared with the traditional image feature based stabilization method,our proposed method gets a much lower complexity as well as a good quality for the output video signals.