基于OpenCL并行加速的Mean Shift三维体数据分割方法

Mean Shift 3D Volume Data Segmentation Method based on OpenCL Parallel Acceleration

Mean Shift算法作为图像分割领域比较经典的算法,是一种基于特征向量的聚类算法,在图像分割的具体实现中应用广泛。为克服Mean Shift算法复杂度高、速度慢的缺点,提出一种三维体数据的快速Mean Shift分割方法。该方法基于Mean Shift的基本思想和GPU的高性能并行计算能力,利用OpenCL对Mean Shift算法进行GPU并行化改造,实现了有意义的分割。实验结果表明,改进后的方法取得了较好的加速效果,运行速度提高了36.44倍。

As a fairly classic algorithm in the field of image segmentation,Mean Shift algorithm is a clustering algorithm based on feature vector,widely used in the implementation of image segmentation.In order to overcome the shortcomings of Mean Shift algorithm,such as high complexity and slow speed,a fast Mean Shift segmentation method for 3D volume data is proposed.Based on the basic idea of Mean Shift and the high-performance parallel computing power of GPU,this method uses OpenCL to transform the Mean Shift algorithm into GPU parallelization,which realizes meaningful segmentation.The experimental results indicate that this modified method could achieve a better acceleration effect and the running speed is increased by 36.44 times.