Local maximum edge cooccurance patterns for image indexing and retrieval
Harpreet Kaur, Vijay Kumar Dhir · 2016
Current research work proposes a new feature descriptor, the local maximum edge co occurrence patterns (LMECoP) for feature extraction for retrieval of images from large database. The LMECoP, firstly collects the local maximum edge information between the referenced pixel and its possible neighbors, then the binary patterns are formed from the extracted maximum edge information. Further the proposed method collects the cooccurrence matrix on the maximum edge information response and to improve the performance of method, the amalgamation of color information and texture information as features is also anticipated in this paper. For color based feature, we convert the RGB color space in to HSV color space and standard histograms are collected for each color space (H, S and V). These color features (HSV histogram) and texture feature (LMECoP) are integrated for the generation final feature vector. The evaluation of LMECoP is performed by tested on Corel-5K database based on average retrieval precision (ARP). The results after investigation it is apparent that the LMECoP outperforms the other existing methods in terms of ARP on Corel-5K database.