An Efficient Method of Image Identification by Combining Image Features

An efficient image identification method by combining image features and using image clustering is proposed. Global and local features in a hierarchical manner are used. The combined global feature reflecting general information of image helps faster retrieval of candidate images and the feature point based local feature facilitates more accurate fine matching with the candidate images. The Fuzzy C-Means clustering method is effective for the image data which are characteristically alike and have fuzzy boundary in coordinate by their global features. As a result, the number of fine matching which requires very large computing time and high complexity is considerably decreased, and matching accuracy is improved.