Research On The Registration And Mosaic Technology Of TDICCD Stitching Images Based On Reflectors

Image mosaic technology is to create a complete and high resolution image with awide field from a group of images which are overlapped with each other by steps ofpretreatment, registration and fusion. In the field of remote sensing image processing,with the development of multi-CCD mosaic space camera, approaches to achieveaccurate registration and mosaic under certain conditions are raising increasingconcern. In this paper, a mosaic method which suits to multi-TDICCD stitching spacecamera based on reflectors is discussed.In the system of TDICCD stitching space camera based on reflectors, vignettingphenomena is existed on the edge of every CCD image plane due to its structure. Twomethods to compensate this phenomena are proposed, Firstly, distribution ofvignetting coefficient on CCD image plane is deduced by analysis from perspective ofgeometrical optics and under this situation, a theoretical model is established tocompensate vignetting. Secondly, functions which fit out all compensation factors forevery pixel under different gray values are created by analysis of radiate calibrationdata. Method using these functions to process the vignetting image is proved toachieve very excellent result.After getting rid of image vignetting, this paper implements as well as presentedan image mosaic algorithm based on an improved Harris feature points detectionalgorithm. Firstly, the improved Harris corner detection algorithm is proposed. Apre-selection method is introduced in this proposed algorithm to improve thealgorithm efficiency and multi-thresholds method is applied to extract corners morereasonably and comprehensively which lay a ground work for the following registration step. Then the detected corner points are coarsely matched using abidirectional registration method based on normalized cross correlation (NCC) afterwhich the Random Sample Consensus(RANSAC) algorithm is applied to remove allmismatching points pairs and solve out parameters of the coordinate transformationmatrix. Finally, after the weighted average method is used to for image fusion, a highresolution mosaic image with wide field is obtained.