Image Retrieval Using Most Similar Highest Priority Principle Based on Fusion of Colour and Texture Features

Abstract

We propose Content-Based Image Retrieval (CBIR) system using local RGB colour and texture features. Firstly, the image is divided into sub-blocks, and then the local features are extracted. Colour is represented by Colour Histogram (CH) and Colour Moment (CM). Texture is obtained by using Gabor filter (Gab) and Local Binary Pattern (LBP). An integrated matching scheme based on Most Similar Highest Priority (MSHP) principle is used to compare the blocks of query and database image. Since each feature extracted from images just characterizes certain aspect of image content, features fusion are necessary to increase the retrieval performance. We present a novel fusion method based on fusing the distance value for each feature instead of the feature itself to avoid the curse of dimensionality. Experimental results in terms of the precision/recall estimates demonstrate that the performance of the proposed fusion method gives better performance than that when either method is used alone.