The Research Of Parallel Image Retrieval System Based On Multi-Core Technology

This article describes a set of content-based image retrieval system, combined with multi-core technology, the system uses multi-threading technology, and give full play to the advantages of the hardware, it realizes the combination of two forms of the image parallel processing technology that are the pipeline in parallel and the data parallel, it improves the system efficiency.The article introduces a number of content-based image retrieval technology, as well as the image in parallel and multi-core technologies, including image color, shape and texture feature extraction algorithm, similarity measure algorithm, the characteristics of two forms of the image parallel processing, which rely on hardware structure and so on.Then the article describes in detail the organizational structure and design of the system, as well as the functional design of the module. The system extracts color moments, Hu invariant moments and the grey level grows matrix of statistical parameters of the image as three sets of eigenvalues of the system, it uses similarity measure architecture of synchronous composition. At the same time, the system in a multi-image input and a small number of input images to provide two manners of multi-threaded approach, the first manner based on shared memory multi-threaded programming model, the allocation of tasks used the method of allocation by region, and in an effort to make the thread assigned to the task of the same amount to the thread load balancing, in order to maximize the speedup. Another way is to start three threads at the same time to extract the image color, shape and texture feature eigenvalues.In this paper, it used dual-core and quad-core computer to experiment, the results showed that with the increase in the number of nuclear CPU, the processing speed of the image retrieval system has linear improvement.