Integrating multi-feature of image based on correspondence analysis

Image feature detection is one of the key techniques of image analysis and image understanding. For a given image, due to the differences in feature extraction methods, the results of feature extraction are not the same. Some methods may results in the loss of certain features of the image, while some may generate extra features of the image. How to integrate the features obtained by different feature extraction methods for a given image to gain a satisfactory result of feature extraction is a very important research topic. In this paper, we employ correspondence analysis theory to integrate image features obtained by using several feature extraction methods for the image. Firstly, the image features are extracted by using different feature extraction methods, and they are arranged as columns to form a data matrix to be analyzed. Secondly, the score of each pixel in the image is calculated by using correspondence analysis for the data matrix. Finally, significance analysis is made of score vectors and the integrated image features are obtained according to the degree of significance. Experiment results illustrate the efficiency of the presented method in this paper.