Research Of Image Matching Based On The Improved RANSAC Algorithm

In many cases, due to a lot of limiting factors, leading to the photos which we have taken, is not the image of our ideal. The limiting factors include:the light source is only a partial light irradiation, the background color of taking photos in the environment is same or similar as the color of the image, the use of camera pixels is doesn't high enough, and the shooting angle is not fit and so on. These factors may cause our image existing extremely dark or bright in local areas. The dark area for extraction of feature points is very difficult, in general, this part of the image, the signal-to-noise ratio is very low, the effective feature points will also be difficult to be detected. Thus, this will cause a large number of false matching pairs appearing when the feature point matching.These are all the problems what we need to solve. In this article, the first to use the bilateral filter combined with retinex method, mainly to enhance the image of the dark area. Use gamma function to correct the enhanced image. To improve image clarity, to increase the resolution, and improve the signal-to-noise ratio at the same time. In addition, the number of feature points which can be extracted also increased significantly. After the stay stitching image enhancement, through the SURF algorithm to extract the feature points and feature point descriptor. And then through the K-D tree to search the nearest neighbor feature point and the second nearer feature points. This step is to follow-up preparing to find the corresponding feature points by the nearest neighbor distance than next near distance matching method, and then constitute a match pairs. Through this matching method can find the matching point quickly and efficiently, Greatly reduces the rate of error matching and improve the matching efficiency. And then, using the improved RANSAC algorithm is proposed in this paper to remove false matching. Through comparing with the traditional RANSAC algorithm, it can be seen that the improved Random sampling method (RANSAC) algorithm can well remove the false matching. At the same time, greatly reduces the time to remove the false matching. Finally, using the weighted average method for image fusion. Through the analysis of experiment, Based on the improved RANSAC algorithm proposed in this paper compared with the traditional RANSAC algorithm to remove miss match, to verify the effectiveness of the improved RANSAC algorithm. And, the improved RANSAC algorithm is more advantageous to the accuracy of the matching and the improvement of the speed of stitching.Simulation results show that through the improved RANSAC algorithm to remove the false matching, witch is proposed in this paper, joining the two images or more of the images witch have the same area of the image together, to become a clear, and wide field of the image.