Image Retrieval by Integrating Global Correlation of Color and Intensity Histograms with Local Texture Features
Suresh Kumar Kanaparthi,U. S. N. Raju,P. Shanmukhi,G. Khyathi Aneesha,Mohammed Ehsan Ur Rahman
DOI: https://doi.org/10.1007/s11042-019-08029-7
IF: 2.577
2019-08-20
Multimedia Tools and Applications
Abstract:Research on Content-Based Image Retrieval is being done to improvise existing methods. Most of the techniques that were proposed use color and texture features independently. In this paper, to get the correspondence between color and texture, we use congruence on Hue, Saturation, and Intensity by using inter-channel voting. Gray Level Co-occurrence Matrix (GLCM) on Diagonally Symmetric Pattern is computed to get texture features of an image. The similarity metrics between two images is computed using congruence and GLCM. To measure the performance; Average Precision Rate (APR), Average Recall Rate (ARR), F-measure, Average Normalized Modified Retrieval Rank (ANMRR) are calculated. In addition to these parameters, one more parameter has been proposed: Total Minimum Retrieval Epoch (TMRE) to calculate the average number of images to be traversed for each query image to get all the images of that category. To validate the performance of the proposed method, it has been applied to six image databases: Corel-1 K, Corel-5 K, Corel-10 K, VisTex, STex, and Color Brodatz. The results of most of the databases show significant improvement.
computer science, information systems, theory & methods,engineering, electrical & electronic, software engineering