Automatic semantic image classification and retrieval based on the weighted feature algorithm

Organizing images into meaningful (semantically) categories using low-level visual features is a challenging and important problem in content-based image retrieval. Clustering algorithms make it possible to represent visual features of images with finite symbols. However, there are two problems in most current image clustering algorithms. One is without considering the choice of the initial cluster centers which have a direct impact on the formation of final clusters, and the other is without considering the relevant features and assigning equal weights to these feature dimensions. According to the two problems we propose a weighted features algorithm. First, we use the labeled image samples to calculate the weight for each feature according to the feature degree of discrete. These weighted features have been used to calculate the initial cluster centers because they can well represent the cluster. Then, we use the weighted features(based on the different image data set) algorithm to discard the irrelevant features and reduce the feature dimensions through the whole clustering process. Experimental results and comparisons are given to illustrate the performance of the new algorithm.