Object detection based on improved color and scale invariant features
Mengyang Chen, Aidong Men, Peng Fan, Bo Yang · Proceedings of SPIE, the International Society for Optical Engineering/Proceedings of SPIE · 2009
A novel object detection method which combines color and scale invariant features is presented in this paper. The detection system mainly adopts the widely used framework of SIFT (Scale Invariant Feature Transform), which consists of both a keypoint detector and descriptor. Although SIFT has some impressive advantages, it is not only computationally expensive, but also vulnerable to color images. To overcome these drawbacks, we employ the local color kernel histograms and Haar Wavelet Responses to enhance the descriptor's distinctiveness and computational efficiency. Extensive experimental evaluations show that the method has better robustness and lower computation costs.