Content Based Image Retrieval Using Full Haar Sectorization

Content based image retrieval (CBIR) deals with retrieval of relevant images from the large imagedatabase. It works on the features of images extracted. In this paper we are using very innovativeidea of sectorization of Full Haar Wavelet transformed images for extracting the features into 4, 8,12 and 16 sectors. The paper proposes two planes to be sectored i.e. Forward plane (Evenplane) and backward plane (Odd plane). Similarity measure is also very essential part of CBIRwhich lets one to find the closeness of the query image with the database images. We have usedtwo similarity measures namely Euclidean distance (ED) and sum of absolute difference (AD).The overall performance of retrieval of the algorithm has been measured by average precisionand recall cross over point and LIRS, LSRR. The paper compares the performance of themethods with respect to type of planes, number of sectors, types of similarity measures andvalues of LIRS and LSRR.