Research And Implementation Of Key Technologies Of Color Image Mosaic

The technology of color image mosaic is the number of overlapping images into a large seamless color image. A complete large seamless picture is gotten by calculating the corresponding relation model between pixels of two adjacent image sequences. So the emphases of this paper are the color image registration, fusion and the seamless mosaic, the specific work as follows:For the SIFT operator in big complex images, extracted many unstable feature points and many false matches when the images are only small overlap regions, which leads to a decrease accuracy in image registration. This paper proposes an improved SIFT algorithm. After extracting SIFT feature from the target image, bidirectional BBF (Best-Bin-First) algorithm is used to match the extracted features, then using the Minimum Neighbor Feature Matching eliminate false matches based on scale and gradient direction information from SIFT descriptor. Using Random Sample Consensus algorithm (RANSAC) further to filter the matching points and calculating transformation model combined with the least square method and polynomial approximation fitting. Finally, the local mean root square error (RMSE) is used to evaluate the mapping error matrix to the actual image. the system will give the feedback and find the error matching, and then iterative to in line with the evaluation. The experimental results show that compared to classical SIFT, SURF and ENVI image registration software, the algorithm of this paper can improve the accuracy of registration, especially for big images with smaller overlap regions.To complete the color image mosaic, after acquiring the same overlap region of image sequence, need to use the image fusion methods integrating the area information of multiple images which being fused. First preprocess the fusion image with camera calibration, affine transformation and exposure compensation by using the information of registration in last step, and then use the adaptive linear mix blend fusion on the overlap region through HIS color space transform. After inverse the transform to RGB color space, use the structural similarity algorithm to calculate the evaluation results of objective evaluation, finally get complete seamless and smooth color image. The experimental results show that compared to the general linear weighted fusion, wavelet fusion method, the algorithm of this paper can eliminate the splicing seam, ghosting of the opera and blur problem more effectively.