COLOR IMAGE RECOGNITION USING VARIANCE-COVARIANCE TECHNIQUE

One of the main problems in pattern recognition and artificial intelligence (Al) is the high computational complexity of searching. In this paper a very simple and efficient algorithm for codebook search is presented The algorithm is based on variance-covariance calculation of 3-D images. In the proposed method, the recognition proceeds by matching individual features to a database of features from known color images using a fast nearest-neighbor algorithm. We apply the full search algorithm to detect the matched image from a database which consists of 300 images. The proposed algorithm employs two test conditions to speedup the test encoding procedure without incurring any extra image distortion. Experimental results on three color image databases are presented. The main advantage of using the proposed algorithm is the high speed in the coding time that reaches 83.08-98.83% depending on the implemented codebook size and the simplicity in calculations