Block DCT coefficients and histogram for image retrieval

Abstract

Retrieving images automatically using low level visual feature is a challenging research topic for the last two decades. Due to the advancement in the digital image technology multimedia content over the internet is increasing every second and thus accurate image retrieval is very important. Color, Texture, Shape and spatial layout are the low level visual features in Content Based Image Retrieval. These low level visual features are used for image representation and retrieval in CBIR. In this paper content based image retrieval system using image features extracted by sub-block based Discrete Cosine Transform(DCT) coefficients in YUV color space and color Histogram in HSV color space is used. Low frequency component of the DCT preserve the most important image features. The block texture feature vector generated from DCT coefficients and color histogram features are trained and classified using Support Vector Machines (SVM). Here a 64 Dimensional DCT coefficient and 32 dimensional histogram is used as feature vector which is more compact hence retrieval is comparatively fast.