Improving the Image Retrieval Performance using False Image Filtering Approach

Abstract

The novel approach combines color and texture features for content based image retrieval (CBIR). This paper is used to retrieve the images from the huge collection of image databases. Most of the research interest in recent years uses feature indexing techniques for the image retrieval. If the number of features are more, then the more time is spent on the comparing the features in low level image retrieval. The proposed system has focused on minimizing the number of comparision by considering the structure of the color theory which says that human color vision system is sensitive to light–dark variations. Here, the color theory is used to eliminate the irrelevant images from the huge collection of images. The feature extraction methods are used to retrive the relevant images. The irrelevant images are filtered by mesuring the deviation between light and dark colors. The opponent values of color and texture features of the image are taken. The image retrieval performance is improved by minimizing the number of comparisions. The proposed method outperforms the other previously developed methods by providing the classification accuracy of more than 89% for the various types of natural images taken from coral database. Hence, this paper concentrates on color and texture features for image retrieval in different directions. The proposed method significantly improves efficiency with less computational complexity.