Image Retrieval with Degree-2 Iterative Linear Interpolation Polynomial Coefficients

Abstract

Images are 2D surfaces. People usually scale images using interpolations, which produce 2D polynomial surfaces. Therefore, in this paper we propose an image retrieval scheme that uses the coefficients of the polynomials as local features. To obtain the coefficients, we apply the iterative linear interpolation (ILI) which produces degree-2 polynomial surfaces. The coefficients are normalized according to the center pixel of the 4x4 sliding window and histogram of the normalized coefficients become the image features for the image retrieval. Experimental results show that for some classes of images, the proposed retrieval outperforms the local binary pattern (LBP), the local ternary pattern (LTP) and the local tetra pattern (LTrP) schemes while it inherits ILI's low computational complexity nature.