A Real-Time SIFT Feature Extraction Algorithm

In order to be implemented for real-time execution and use color information, an improved scale-invariant feature transform (SIFT) algorithm based on the contrast enhancement and DAISY descriptor is proposed in this paper. The color offset and the exposure offset of the pixel are calculated from the color image and the original grayscale image. The two offsets are added to the gray level to gain the contrast enhancing image according to which the SIFT feature points and the main directions are got. The DAISY descriptor of the feature point is obtained from the partial derivative values of the feature point and the sampling points with different directions. To evaluate the accuracy of the algorithm under different view changes, the feature points extracted from the standard image data sets are matched by the K-nearest neighbor (KNN) algorithm. The experimental results show that the algorithm can obtain more feature points. Furthermore, it guarantees the accuracy while reducing the time of the descriptor generating process and the feature matching process.