Accelerating of Color Moments and Texture Features Extraction Using GPU Based Parallel Computing

Image retrieval tools can assist people in makingefficient use of digital image collections; also it has becomeimperative to find efficient methods for the retrieval of theseimages. Most image processing algorithms are inherentlyparallel, so multithreading processors are suitable in suchapplications. In very big image databases, image processing takesvery long time for run on a single core processor because of singlethread execution of algorithms. GPU is more common in mostimage processing applications due to multithread execution ofalgorithms, programmability and low cost. In this paper weimplement color moments and texture based image retrieval(entropy, standard deviation and local range) in parallel usingCUDA programming model to run on GPUs. These features areapplied to search images from a database which are similar to aquery image. We evaluated our retrieval system using recall,precision, and average precision measures. Experimental resultsshowed that parallel implementation led to an average speed upof 144.67×over the serial implementation when running on aNVIDIA GPU GeForce GT610M. Also the average precision andthe average recall of proposed method are 61.968% and 55%respectively.

Keywords:

color moment , CUDA , GPU , texture based image retrieval