Abstract:

The color scaleinvariant feature transform （color SIFT） based feature extraction method is proposed, and the related invariance properties in translation, rotation, zooming, and color shifting are analyzed. Further, image classification strategy and algorithms based on this feature are studied. To test the proposed classification and recognition algorithms, 50 objects categories randomly chosen from the Amsterdam library of object images (ALOI) are employed for recognition, and results show that the correct rate of recognition is 100%. Both theoretical and experimental results validate that the color SIFT feature has a good performance in chromatic image classification and recognition.