【摘要】 在图像处理和模式识别中,通过细化算法抽取图像的骨架是一项极有意义的工作。由于大多数传统细化算法都需要按照一定顺序对每个像素进行逐点扫描,执行效率不高,另外图像细化算法中数据之间的关联性不大,将传统的快速并行细化算法移植到CUDA(compute unified device architecture)平台上,运用CUDA的并行处理优势对图像进行细化处理。实验表明,该方法在保证细化处理效果的前提下,还能将运算效率提高达到幂级数级。更多还原

【Abstract】 In the image processing or the pattern recognition,extracting the image skeleton through thinning algorithm is a very meaningful work.As most of the traditional algorithms need to scan the pixel point by point with its poor efficiency,and for another way,the connection between image data is weak,so this paper transplant the thinning algorithm to CUDA platform and run with its parallel computing.Experiments show that the operation efficiency is improved under the precondition of the high precision of the image.更多还原