简单线性迭代聚类的超像素算法

A Superpixel Algorithm Based on Simple Linear Iterative Clustering

当前,图像分割处理中存在过程繁琐、效率低下等问题,加上图像超像素形状及大小不均等影响因素,难以适应高性能计算要求。基于此,本文提出了简单线性迭代聚类改进算法。首先预处理获得像素信息,减少算法执行过程中的计算耗时;其次,通过一系列算法优化,降低访存时间,使其更有利于后续并行化的操作;最后依据像素点的处理之间互不产生依赖的情况,使用向量化和并行化的手段,对算法进行整体加速,进一步提升算法效率。实验结果表明,本文提出SLIC改进算法相比基准模型,在计算速率和算法性能方面得到提升。

Nowadays,there is a cumbersome and inefficient process in image segmentation processing,and it is difficult to adapt to the requirements of high-performance computing due to the unequal shape and size of image superpixels.Based on this,this paper proposes a simple linear iterative clustering(SLIC)improved algorithm.First of all,the pixel information is preprocessed to reduce the calculation time in the process of algorithm execution;Secondly,through a series of algorithm optimizations,the access time is reduced,and it is more conducive to subsequent parallelization operations;Finally,according to the situation that there is no dependence between the processing of pixels,vectorization and parallelization are used to accelerate the algorithm as a whole to further improve the efficiency of the algorithm.Experimental results show that compared with the benchmark model,the SLIC improved algorithm is proposed to improve the calculation rate and algorithm performance.