A Comparative Study of Sift and PCA for Content Based Image Retrieval

This paper presents a comparative approach for Content Based Image Retrieval (CBIR) using Scale Invariant Feature Transform (SIFT) algorithm and Principal Component Analysis (PCA) for color images. The motivation to use SIFT algorithm for CBIR is due to the fact that SIFT is invariant to scale, rotation and translation as well as partially invariant to affine distortion and illumination changes. Inspired by these facts, the paper investigates the fundamental properties of SIFT for robust CBIR by using binary MPEG-7 and Grayscale COIL-20 and Color image databases. Our approach uses detected keypoints and its descriptors to match between the query image and images from the database. Our experimental results show that the proposed CBIR using SIFT algorithm produces excellent retrieval result for images with many corners (concaves) and edges (convex) as compared to retrieving image with less corners and edges. The paper also presents another approach for CBIR using Principal Component Analysis (PCA) for Color images. The main aim of the paper is to employ SIFT and PCA methods on the same Image databases and perform a comparative study of results between SIFT and PCA approaches using Precision and Recall tables. The study reveals that SIFT approach provides a better Image retrieval performance for binary, gray scale and color images when compared to PCA.