Objects Extraction Based-Disparity And Stereoscopic Video DT Mesh Compression Coding

Stereoscopic video display technology is bound to get more development in themultimedia field with the rapid development of3D technologies. The prospect ofstereo video technology application is more and more broad currently. Stereo video isdesigned to emulate human stereo perception, scene depth information of stereoscopicvideo enhance the visual sense reality and realism. However, one major obstacle forthe application of stereo video is the extremely large amount of data, which broughtgreat difficulties to the transmission of video data, Therefore, the stereoscopic videocoding is increasingly important, which has aroused extensive attention of numerousresearchers. This paper mainly works on the stereoscopic video coding based-object orbased-Delaunay Triangle mesh.Considering the disparity information of stereoscopic video and motioninformation of continuous frames, a stereoscopic video object extraction algorithm isrecommended in this paper. First, In order to improve the accuracy and the speed ofstereo matching, reprocess and reduce sampling the left and right channels ofstereoscopic video images respectively, Obtain the disparity maps utilizing the one-dimensional windows of stereo matching for the two channels respectively, and extractthe the disparity maps after median filtering for the first time, approximate videoobjects are obtained from the extraction of disparity map, getting the binary extractionmasks and post-processing it, recovery the size of the original image by expandsampling. On this basis, detect the movement region of the same disparity-levelcombining with the symmetric difference algorithm, At last, obtaining the accuratevideo object mask, and extract the final video objects utilizing the masks.Based on the extracted video object, considering the sensitivity of the human eyeson the video objects is greater than the background, a Based-object stereoscopic videocoding algorithm is proposed. The algorithm depends on the video encodingframework of the H.264standards, the video objects and background objects will be encoded adopting different quantization steps in the left channel, For the right channel,The disparity information which is between the left and right sequences of stereoscopicvideo will to be coded only, instead of encoding the right channel, However, all thedisparities contain objects disparity and background disparity, so adopt differentquantization methods for two different disparities. Reconstructed images of the rightchannel can be got from there constructed images and the correspondingdisparity. Taking account of the disparity information of the stereoscopic video and thedifferent sensitivity of the human eyes with different regions, the algorithm willimprove the compression ratio and guarantee the visual images.This paper proposes a based-DT mesh representation method for the disparity mapof stereo images, which is designed to concentrate on stereo image compression. Firstconstruct DT mesh adopting the characteristic points that express disparity map, andmap it to the stereoscopic video image, then disparity compensation with theredundancy information between the left and right channels. Motion estimation andcompensation MB-SM affine transformation of the mesh nodes, finally encode thedisparity information and motion information, thus achieve the goal of stereoscopicvideo coding.The proposed method for the disparity maps estimation and motion estimation, whichabandon the utilization of block-based matching method. The experimental resultsshow that the proposed method has better performance in stereo image compressionthan the conventional methods, avoiding the block effect, taking accountof the disparity information of stereo video, so the reconstructed image has greatercorrelation and stronger stereoscopic visual perception.