A Perceptual Model Based on Computational Features for Texture Representation and Retrieval

A new perception model based on image representation and retrieval for a set of computational measures is proposed in this paper. We consider a set of textural features that individuals use to identify and categorize textures having a perceptual meaning and their application to content-based image retrieval. Such features include coarseness, directionality, contrast, and busyness. This paper proposed a new method to calculate a set of perceptual texture features. The perceptual model presented is judged using a psychometric method (based on rank-correlation) and found to represent very well to human judgements. For these measures large database is required. Therefore the Brodatz database and benchmarking based on exploratory results gives exciting performance. This paper proposes to use two representations for better retrieval efficiency: the original image representation and the autocorrelation function representation. In this paper with the help of autocorrelation function related images are presented to the given input image (based on texture and colour). The related images are displayed either the user satisfies or until no change. The compatibility of the preferred computational measures is shown by human judgement. Firstly, based on the spearman rank-correlation coefficient. Second, the proposed computational measures in texture retrieval shows exciting results and their application mostly when using results returned by each of two representations.