Improving the Feature Saliency by Eigenvalue Analysis

Human visual system can improve the feature saliency automatically from insufficient contrast or when there are occlusions . This paper focuses on corner/junction detection by simulating the biological vision mechanisms to increase the feature saliency. An algorithm is designed to simulate the recognization process of biological vision system. This approach depends on the orientations of edges. The feature saliency measure are increased from the image. The algorithm has been applied on several classical contour test images and the experiments demonstrate that the algorithm is more simple and can handle junctions robustly compared with previous works .