Research And Realization On Parallel Curve Tracing Algorithm Of Image

With the rapid development of computer technology, Image processing algorithm in the application of social life play an irreplaceable role. The key technology curve tracing of image as image processing and pattern recognition, plays a more and more important role in engineering, medicine, industrial inspection and road traffic etc. many other fields. Curve tracing in this paper refers to the process of extracting all the curves in the image coordinates of points by order.Due to the explosive growth of data quantity and industry on the algorithm to deal with the real-time need, research on the algorithm of acceleration methods has become a hot research field. NVIDIA launched the CUDA parallel computing programming framework model of accelerated algorithm provides a good technical platform, which makes the GPU universal parallel computing has become the mainstream. The realization of curve tracing algorithm parallel accelerate achieved good real-time performance.This paper mainly studies the image curve current extraction related algorithms, based on the actual project, proposes a new adaptive parallel curve tracing algorithm of image based on CUDA. The algorithm mainly consists of three parts:threshold segmentation, image thinning processing and image curve ordering extraction. Threshold segmentation and image thinning as the image preprocessing, get the single pixel width of two value image. After curve ordering extraction algorithm, finally obtain all possible curve set.This paper presents a novel curve tracing algorithm and achieved base on CUDA for parallel acceleration. Realize three parallel threshold segmentation methods:the threshold specified method; multi threshold selection method; automatic optimal threshold method. Realize parallel thinning algorithm based on index table. Put forward image curve extracting parallel algorithm, give full consideration to the trend of curve in complex situations, extraction is realized orderly curve pixel coordinates with the help of graph theory technology. Utilize memory optimization, instruction and program code optimization, concurrent optimization strategy for algorithm acceleration. Finally verified and analysis by experiments, tracing algorithm has high extraction efficiency, good treatment efficiency on intersection and curve breakpoint continuous and other good characteristics.