Research On Color Correction And Image Fusion In Image Stitching

According to statistics, more than80%information accessed by human beings through seeing. As we all know, video and image are two extremely important carriers of information which is obtained by people. Therefore, to get and process video or images becomes more and more important. Especially, with the development of science and technology in recent years, digital camera becomes more and more universal. The need of processing video and image is rapidly increasing, whether military area or civilian area. In civilian area, such as, get panoramic images or video. In military area, such as, monitor the enemies, and so on.This thesis selects two key technologies in video or image stitching process as the research points:The first one, to do color correction of the source images. Because the image’s color is vulnerable to the external environment and the parameter settings of image capture device. Therefore, even use the same camera to capture the same scene under different illuminations, the measured color will be slightly different. The second one, to merge aligned images, this step will also directly affect the final stitching effect, such as whether there will be a ghost, and whether there will be splicing gap.For the first point, we propose a color correction which is based on local image blocks and this is different with the previous color correction methods based on the entire image in order to get as accurate results as possible. The change of illumination on images is uneven and using the entire image to calculate conversion parameters will get the result with large error. But if we break the entire image into many small blocks, we can consider the change of illumination of the block is even.For the second point, we propose a hierarchical fusion algorithm and the final result will be obtained by accumulation. Firstly, we compute the Gaussian pyramid and Laplacian pyramid of the source images respectively. Secondly, we fuse on Laplacian pyramid and then accumulate all the levels of Laplacian pyramid from top to bottom. There are32pictures,4tables,65references.