Implementation Of Hough Transform Parallel Algorithms Based On CUDA Platform

Feature Extraction, has always been an important research direction in the field of digital image processing, and the Hough Transform is belongs to the category of extracting the edge feature of the image. It is based on the duality of the image space and the parameter space, and clusters the pixels which have a certain relationship in the image. Just because of the capability of resisting disturbance, and cannot be affected by the shape defect, the Hough Transform has been widely used in engineering and medicine field.For the past few years, parallel computing becomes the trend of development, among the various platforms for parallel computing, the compute unified device architecture CUDA based on GPU designed by NVIDIA has received widely attention, due to it has took full advantages of the computing capacity of GPU.This paper put forward a line and circle detection algorithm based on CUDA architecture, according to the characteristic that Hough Transform is suitable for parallel computing. And the experimental results show that, compare to the other similar algorithm, this one has a certain extent lifting in the characteristic.Firstly, this paper gives an introduction of the parallel computation and feature extraction, secondly, illustrates the programming model, software and hardware architecture and memory hierachy in detail. Then introduces the principle of line and circle detection algorithms using Hough Transform, designs a parallel algorithm of line and circle detection using Hough Transform for CUDA, and complete the design through the CUDA hardware platform, and then does some performance optimization for this algorithm through a series of strategies. At last, this paper makes a detailed experiment for the line and circle detection based on Hough, proves that the Hough Transform has the characteristic of resisting disturbance and the advantages of CUDA in the field of massively parallel computing.