A SIFT-Color Moments Descriptor for Object Recognition

Feature extraction technique has been widely studied and used in many fields, such as Augmented Reality, 3D Reconstruction and object recognition. In recent years, intensity-based descriptor have been widely used for feature extraction, and the SIFT descriptor is the most robust of them. However the color information is not included in SIFT, and the color provides important information in object description and matching tasks. SIFT can´t differentiate the objects with similar shape but with different colors commendably. Many objects can be misclassified in object recognition without color information. Therefore, this paper proposes a novel descriptor combine SIFT with Color Moments to improve the performance of object recognition, and so called SIFT-Color Moments Descriptor. Experimental results show that the SIFT-Color Moments Descriptor is more robust than the traditional SIFT with color image.