MULTILEVEL APPROACH OF CBIR TECHNIQUES FOR VEGETABLE CLASSIFICATION USING HYBRID IMAGE FEATURES

Abstract

CBIR is a technique to retrieve images semantically relevant to query image from an image database. The challenge in CBIR is to develop a method that should increase the retrieval accuracy and reduce the retrieval time. In order to improve the retrieval accuracy and runtime, a multilevel CBIR approach is proposed in this paper. In the first level, the color attributes like mean and standard deviations are proposed to calculate on HSV color space to retrieve the images with minimum disparity distance from the database. In order to minimize search area, in the second level Local Ternary Pattern is proposed on images which were selected from the first level. Experimental results and comparisons demonstrate the superiority of the proposed approach.

Keywords

• Content Based Image Retrieval (CBIR)
• Gray Level Co-occurrence Matrix (GLCM)
• Local Binary Patterns (LBP)
• Local Ternary Pattern (LTP)