Pruning and Weighting of Keypoints Using the HSI Color Space for Image Recognition

The detection of the stable local image features is one of the most critical tasks for many object recognition algorithms. The Scale Invariant Feature Transform (SIFT) has been shown to be effective to the image matching or object recognition. However, the large number of features generated by the SIFT is a disadvantage for the real time application. In this paper, we present a novel approach to detect more important local features by finding the higher information keypoints (HIKs). An input color image is firstly decomposed into an intensity image, a hue image and a saturation image. Then we detect the HIKs in these color component images in terms of the keypoint positions. Furthermore, a weight for each HIK is assigned according to the position relationship of the keypoints to improve the matching accuracy. Experiments show that the proposed approach can achieve higher matching accuracy and reduce the matching time by using the HIKs and their weights.