Transformed Directional Tri Concomitant Triplet Patterns for Image Retrieval

Abstract

Content-based image retrieval (CBIR) is an accurate characterization of visual information. In this paper, we have proposed a new technique entitled as Transformed Directional Tri Concomitant Triplet Patterns (TdtCTp) for CBIR. TdtCTp consist of three stages to obtain detail directional information about pixel progression. In first stage, structural rule based approach is proposed to extract directional information in various direction. Further, in second stage, microscopic information and correlation between each sub-structural elements are extracted by using concomitant conditions. Finally, minute directional intensity variation information and correlation between the sub-structural elements are extracted by integrating first two stages. Retrieval accuracy is estimated using traditional distance measure technique in terms of average retrieval precision and average retrieval rate on publicly available natural and medical image databases. Performance analysis shows that TdtCTp descriptor outperforms the existing methods in terms of retrieval accuracy.