A novel content-based image indexing and retrieval framework using clockwise local difference binary pattern (CWLDBP)

Abstract

In this chapter, we propose a novel content-based image indexing and retrieval framework based on clockwise local difference binary pattern. Local binary pattern (LBP) is a popular texture content-based image indexing and retrieval framework proposed by Ojala et al. [1]. In our proposed method, we find the pair-wise difference between the pixels in clockwise direction of 3 × 3 neighborhood and the result of difference is used to find the binary pattern. Three novel methods such as CWLDBP1, CWLDBP2, and CWLDBP3 are proposed. The main advantage of the proposed methods such as CWLDBP1 and CWLDBP2 is the use of few binary patterns compared to LBP. That is, our proposed methods CWLDBP1 and CWLDBP2 use only sixteen binary patterns to index an image. To test our proposed method, we use Corel 1k database. Our proposed method has shown reasonable results when compared to 3 × 3 neighborhood-based LBP rotational invariant, LBP uniform, and LBP uniform rotational invariant methods.