THE COMPLETE RANK POSITION MATRIX FOR CONTENT BASED IMAGE RETRIEVAL
G. Bindu Madhavi, V. Vijaya Kumar, Kadiyam Sasidhar · Journal of Critical Reviews · 2020
This paper derives a new local based frame work for content based image retrieval (CBIR). This paper initially computes advanced symmetric local binary pattern (AS-LBP) and replaces the center pixel of the 3x3 window with AS-LBP unit. The AS-LBP is different from symmetric LBP. The AS-LBP considers the weighted central pixel in deriving symmetric relations of sample pixels and also considered more edge information in the form of cross and diagonal differences, etc. The AS-LBP unit image is divided into 2x2 grids and on each grid this research derives complete rank sequences. This research created a complete rank sequence matrix (CRSM) which contains all possible rank sequences. The 2x2 grid is replaced with compete rank position (CRP) and the gray level co-occurrence matrix (GLCM) features on CRPM image are used as feature vector for CBIR. The proposed CRP and CRPM are more powerful than ordinary rank sequence. The ordinary rank sequence does not allow repetitive rank scores. The CRPM frame work is experimented on wide range of databases and compared with the state-of-art methods.