Interest point detect method to stereo images with good repeatability and information content

For the purpose of stereo image matching, several common stereo image interest point detectors are reviewed in the two aspects of repeatability and information content through experimental analysis. Based on the image information entropy, clustering is utilized to analyze the texture features of images, and an interest point detect method is then proposed. The most significant characteristic of this method, which improves the Harris detector, is self-adaptive to the image texture. The experiments by using of actual stereo images with different texture features proved that this method guarantees more better repeatability and information content.