基于GPU并行计算的图像二值化技术和研究【毕业论文】.doc

本科毕业论文 （20 届） 基于GPU并行计算的图像二值化技术和研究 专业：计算机科学与技术 摘要 二值化是将一个图像转变成为二值图像，即只有0和255两种亮度值。二值化的关键是阈值的选取，通常大于阈值的就是255，反之则设定为0。在数字图像处理中，二值图像占有非常重要的地位，图像的二值化有利于图像的进一步处理，使图像变得简单，而且数据量减小，能凸显出感兴趣的目标的轮廓。 The Study Based on GPU Parallel Computing and Technology of Image Binarization Abstract Binaryzation is a way to make an image into a binary image, that only 0 and 255 two brightness values. The key of binaryzation is threshold selection, Usually turn the greater than the threshold into 255,conversely set to 0.In digital image processing, binary image is occupied a very important position, The binaryzation of image is benefit for its further processing. Make the image become simple, And decrease the data quantity, Can highlight interesting outline of the goals. Image binaryzation is the premise of image pattern recognition, Image recognition directly determine the result of identification. Due to the need for effective binary image of each pixel to determine the binary values, So the amount of calculation is very large and Hard to achieve real-time requirements, binary needed to effectively improve. CUDA is The model of NVIDIAs GPGPU, it can directly to use most C which people are familiar with ,to write execution procedures in display chip. As a result of current display chip has the ability of highly programmable, with The high memory bandwidth, and a large amount of execution unit, compared with high order CPU, the price of Video card is relatively cheaper. The framework of CUDA programming is more simple than traditional programming, suit for solving the problem can be large parallelized, And for the problem is not parallel processing can be completed by the CPU, Carry out the linkage between CPU and GPU, improves the efficiency of Computation , Reduces the consuming of computational time. Experiments show, compared with the traditional binaryzation based on CPU, Use the image CUDA binary technology can reduce about 90% of calculation time. 【 Keywords 】binaryzation, CUDAGPU