On Fusion And Stitching Algorithm For Panoramic Video Image

Virtual Reality has achieved a major leap in last five years in popularizing technology and expanding market.In particular,commercial applications of VR 360 panoramic video have showed up in thousands of families and entered a period of rapid development,the increasing demand for subjective quality and immersive feelings leads to a fast-rising amount of panoramic video data with high-resolution and high-dynamic ranges.That directly makes VR video data simultaneously facing greater challenges in the aspects of transmission and storage.Therefore,it is a strongly reality urgency that study VR panorama video synthesis and coding compression.This paper focuses on fusion,stitching and some technologies related to coding in panoramic video image,and specifically works in four aspects.(1)Aim to achieve the whole process of panoramic video image with acquisition,pre-processing,stitching and fusion.Based on the video data collected by the hexahedral six lens camera,using checkerboard calibration method to calibrate the camera,adopting the white balance algorithm to solve the consistency problem of image chromatic aberration.The work maps six images to the sphere,stitch and fusions them via SIFT algorithm and the method of overlapping region linear transition,respectively,then finally the VR360 panoramic video image is generated.(2)Select the best input format of panoramic video encoding among various mapping formats,which from VR sphere view to the plan one.The work analyzes and compares the mapping transformation methods,such as cylindrical,hexahedral,octahedral,icosahedron and new tile segmentation,to find out their differences in coding performance,subjective-objective quality and computational complexity.Through verification,it conclude that hexahedron is currently the best coding input format.(3)The technology based on image fusion and stitching provides a cross-layer folding search algorithm for panoramic video motion estimation.Aim to reduce the error of the motion estimation search block and improve the coding efficiency under motion compensation,the work splices and merges the reconstructed left and right borders of frame by the image stitching technique,in use of continuity of the left and right borders of the panoramic video image.(4)To propose a removal algorithm of grating fringe noise for image of CIS scanners.The work makes analysis of physical structure of the CIS scanner,establishes the mathematic model of the grating fringe noise and puts forward a Fourier function algorithm.These steps can effectively remove the fringe noise of the image.The work of four aspects above is expected to provide some useful reference for algorithm research about panoramic video fusion,stitching and coding.In the aspect of panorama image fusion and stitching algorithm,the work synthetize panoramic image smoothly and naturally,and get a good subjective quality.The result about experimenting various mapping format from VR sphere view to plan one,shows that the hexahedron is the existing mapping scheme which is superior to the latitude and longitude map,and it is recommended as the main choice of the panoramic video coding input format.The estimation folding algorithm of panoramic video motion can find a smaller matching block of distortion,effectively improving the coding efficiency.The grating fringe noise removal algorithm can effectively remove the CIS scan image noise and is better than the existing noise removal algorithm.