The Research On Image Filtering Technology Based On CUDA

June2007NVIDIA launched CUDA (Compute Unified Device Architecture)solves GPU general computing usability problems． It does not need to master theknowledge and the corresponding computer graphics API interface, programminglanguage based on the C language and it can be extended so that it could reduce theprogramming complexity．With real-time image and efficiency of information processing requirements arehigh; the amount of data to be processed largely involved in the calculation process．Computing performance can’t meet the performance requirements of practicalapplications． The use of GPU accelerated computing can obtain an order ofmagnitude faster, so GPU general computing become a hot research． How to usepowerful computing and parallel of GPU to accelerate image processing capabilityis important for real-time image information and efficient processing．Based on the study of the existed CUDA-based platform to expedite theprocessing of image-based technology, for the image processing part of the filter,this article presents a method that improve the efficiency of image processingthrough CUDA and OpenGL interoperability． In this paper, the filtering algorithmand the CUDA parallel programming model will be combined, using CUDA andOpenGL interoperability to accelerate image display． Its advantage lies throughDMA (direct memory access) to read and write on the graphics pixel data quicklywithout involving the CPU cycles． Image data is loaded in the pixel buffer andtransferred from the pixel to the texture buffer zone both completed by the GPU．Meanwhile OpenGL allows asynchronous DMA transfer, CPU does not have to waituntil all the pixel data transfer is complete that it can go to perform other tasks, sothat largely reduce the time image processing and improve the efficiency of imageprocessing． Finally, the different size of the image is processed and the resultproved that with the amount of data increases, the GPU acceleration will get goodresults．...