Improving the Point Correspondence Accuracy of Kanade-Lucas Method by Affine Transformation and Correlation

Abstract

The Kanade-Lucas (KL) feature tracker is one of the better matching methods. However it fails to find corresponding points of two images when image motion is quite large. We propose the feature point matching method, making use of affine transformation and correlation so that the KL method works well for large image motion. First, the KL method is used for matching points. Some points are matched well, other points are mismatched or failed to be matched. Second, correlation of two matched points between the first and the second image obtained by the KL method is used to discriminate between well matched points and badly matched points. If the correlation coefficient is lower than a specified threshold, it is regarded as an outlier. Third, all the points to be matched in the first image are transformed on the second image by affine transformation of well matched points alone. However, all the transformed points from the first image to the second image are not well matched well due to the movement of the image. The role of affine transformation is to reduce the computation time and the number of mismatching points by designating only a small searching area around the transformed point. Finally, correlation is used to find the best matching point around the transformed point. Real image data has been used to test the proposed method, and excellent results have been obtained with an average error of 0.811 pixels.