An Efficient Content Based Image Retrieval using Statistical Soft Computing and Texture Features

With the invent of low cost cameras the uses of imaging data has exponentially increased in last two decades. Due to availability of huge data on web, demand of efficient image retrieval techniques have also increased. Many feature based local and global methods have been designed in past but they were either too complex or only case specific. In this paper a simple and efficient statistical soft computing and texture based content based retrieval system is proposed and designed. The method is designed to match the quarry and template images based on histogram and their statistical properties as statistical absolute mean difference and 2D normalized correlation of texture images. Method first resizes the quarry and template image to same size and then calculates the statistical parameters in RGB domain and compares the same. In addition Local binary pattern (LBP) is calculated for comparing the local texture feature of the quarry and template images. The performance of our proposed method is tested and evaluated using the standard large image-vary dataset of color images.