Research On Image Stitching Technology With Parallax Robustness

With the development of image information technology, the application of image stitching technology has become more and more extensive. After research, scholars have put forward a lot of excellent stitching algorithm. Traditional image stitching technologies have certain requirements for scenes and shooting conditions: the scene can be approximated as a plane or images gained in the same viewpoint. And for the images with parallax gained in different viewpoints, the result is not ideal, prone to false image and object error and so on. This paper researches the stitching algorithm for images with parallax gained in different viewpoints, and proposed a stitching algorithm with parallax robustness based on the traditional stitching framework. The main research works are as follows:(1) We proposed a feature point matching pair filter method based on plane similarity. For images with parallax, there may be several planes in the scene. According to the characteristics that point matching pairs in one plane meet with a transformation matrix, and the matrixes calculated from point matching pairs in different plane may be quite different, we define the planar similarity between the feature point matching pairs, and design matching pair filter algorithm, to select the correct matching feature point matching pairs. Effectively solve the problem of disparity image, and improve the registration accuracy.(2) In light of the situation that there may be color difference between different images according to the difference of the shooting pattern or illumination, we define a color difference between images according to the RGB value of the overlap region. We use the extended Gamma correction algorithm to correct the color of image with large color difference. The advantage of this method is that it is able to automatically to correct the color difference image by setting the difference threshold.(3) Combined with the above methods, the design and implementation of the with the robustness of the parallax image mosaic algorithm, and compared with the other image mosaic method, experimental results prove the algorithm in this paper of parallax images can also very good stitching, has certain disparity robustness. We apply this stitching algorithm to the video monitoring system, so that the monitoring personnel can easily observe the entire monitoring scene. And we divide the whole stitching algorithm into two parts: initialization and splicing, which basically meet the real-time requirements of the actual system.