Research On Stereo Video Image Compression

In the process of the power line inspection with the Unmanned Aerial Vehicles(UAV), space information from stereo video camera is more conducive to determine the status of transmission lines and more realistic reproduction of natural scenes of transmission lines. To real-time stereo video signal transmission, the bandwidth is required to achieve the dual-channel transmission requirements. In order to reduce the bandwidth, it is necessary to study compression method of the stereo video signal. In this dissertation, the algorithm of stereo video compression was studied based on morphological wavelet transform theory and certain results were achieved.Disparity estimation based on morphological wavelet transform and Delaunay triangular mesh, as well as image compression based on disparity embedding were addressed. First, the morphological wavelet was used to obtain the characteristic points of the reference image to generate a Delaunay triangular mesh only for foreground. And then, the characteristic points were taken as the center for disparity estimation in the target image to get the disparity vectors of the characteristic points. Finally, the disparity map was generated by using interpolation, after which, the left image were embedded with the disparity vectors and compressed. The experimental results show that the method can effectively extract the edge information, obtain a more accurate smooth disparity field, and improve the efficiency of data encoding and the compression ratio by eliminating the disparity map coding.Based on the actual shooting images of the power lines, the characteristics of the power line image were summarized, and the effectiveness of the proposed method for the power line stereo image compression was verified. Finally, based on the foregoing summary steps of stereo image compression was synthesized, a flow diagram of stereo image compression system is showed, and the system implementation process of the image acquisition from a remote terminal to the local three-dimensional reconstruction is simulated and displayed.