Efficient Image Recognition Using Local Feature and Fuzzy Triangular Number Based Similarity Measures

Image local scale invariant features are of great importance for object recognition. Among various local scale invariant feature descriptors, Scale Invariant Feature Transform (SIFT) descriptor has been shown to be the most descriptive one and thus widely applied to image retrieval, object recognition and computer vision. By SIFT descriptor, an image may be described by hundreds of key points with each point depicted by a 128-element feature vector; this representation makes the subsequent feature matching very computationally demanding. In this paper, we propose to incorporate the fuzzy set concepts into SIFT features and define fuzzy similarity between images. The proposed approach is applied to image recognition. Experimental results with the coil-100 image database are provided to show the superiority of the proposed approach.