距离变换在图像处理中有着非常广泛的应用。由于3D图像数据的复杂性,传统基于CPU的3D距离变换效率较低。为此,研究了将3D图像数据有效地组织到纹理中存储的方法,设计并实现了基于GPU的3D距离变换并行算法。实验结果表明,相对基于CPU的算法,该方法具有非常高的加速比。
Distance transform had gained numerous applications in image processing. Traditional CPU-based 3D distance transform was inefficient because of the complexity of 3D images. This paper proposed an efficient method to store 3D images in the texture of graphics hardware. A parallel 3D distance transform algorithm based on GPU was implemented. Experiments show that the proposed GPU-based method gain very high speedup ratio compared with CPU-based algorithm.