Co-occurrence of Edges and Valleys with Support Vector Machine for Content Based Image retrieval
M. Bennet Rajesh, Sarmitha Sathiamoorthy · 2020
Over the decade, image retrieval attracted many researchers owing to the enormous increase in digital images and digital image repositories. In this paper, co-occurrences of BDIP (block difference of inverse probabilities) is employed as a shape feature which extracts edges with valleys more effectively. After computing the BDIPs, GLCM has applied over the BDIPs of the image along with four directions and it results in four co-occurrence matrices of BDIPs and is combined for the proposed image retrieval. SVM is employed in the classification phase to significantly increase the accuracy and decrease the time cost of the proposed system. The proposed system is tested on Holiday, Corel 10k and Caltech 101 datasets and results evident that proposed system obtained better efficiency then state-of-the-art methods.