Mahidol University Logo
Faculty of ICT, Mahidol University
 

Admissions

Printable Version

 

COMPARISON OF COLOR TEXTURE FEATURES BASED ON PARALLEL AND INTEGRATIVE APPROACHES FOR IMAGE RETRIEVAL

 

TITLE COMPARISON OF COLOR TEXTURE FEATURES BASED ON PARALLEL AND INTEGRATIVE APPROACHES FOR IMAGE RETRIEVAL
AUTHOR CHAREERAT PRACHAKTHAM
DEGREE MASTER OF SCIENCE PROGRAMME IN COMPUTER SCIENCE
FACULTY FACULTY OF SCIENCE
ADVISOR SUKANYA PHONGSUPHAP
CO-ADVISOR DAMRAS WONGSAWANG
 
ABSTRACT
Color feature and texture feature are widely used for image analysis and image retrieval. Combining color and texture features together should enhance retrieval performance. Two combining approaches were studied; parallel and integrative. Two versions of the parallel approach were employed, derived from concatenating co-occurrence matrices with HSV color histogram, and Gabor Filter based features with HSV color histogram, respectively. The integrative approach can be further differentiated into a single channel and a multi-channel method. The single channel method applies texture feature in each color channel separately, using both RGB and HSV color models. The multi-channel method handles two or more channels simultaneously. The multi-channel integrative techniques used in this research were the 3D co-occurrence matrix, and the complex color Gabor Filter. For similarity measurement, were adopted L-1 norm, L-2 norm, L-8 norm, and normalized Euclidean distance. In the image retrieval experiments, we used 291 categories of ceramic tile images, each category consisting of 6 images totalling 1,746 images. Results indicate that for the parallel approach, Gabor features concatenating with 224-bin color histogram with L-8 norm yielded the highest average precision (90.73%). For the single channel integrative approach, Gabor Filter in three channels of R, G, and B with L-8 norm gave the highest average precision (96.34%). Complex color Gabor Filter with L-8 norm achieved the highest average precision (83.51%) for the multichannel integrative approach. According to the experimental results Gabor Filter in three channels of R, G, and B is the best color texture feature for ceramic tile image retrieval system.
KEYWORD COLOR IMAGE RETRIEVAL, CO-OCCURRENCE MATRIX , GABOR FILTER BASED FEATURES, COLOR TEXTURE FEATURES

 

Go to Top

 

ICT Building, Mahidol University, 999 Phuttamonthon 4 Road, Salaya, Nakhonpathom 73170 Tel. +66 02 441-0909 Fax. +66 02 849-6099
Mahidol University Computing Center, The Faculty of ICT, Mahidol University , Rama 6 Road, Rajathevi, Bangkok 10400 Tel. +66 02 354-4333 Fax. +66 02 354-7333