基于类间蚂蚁竞争模型的显著图像分割算法

Saliency Image Segmentation Algorithm Based on Ants Competition Model

由于单类蚁群算法分割易造成欠分割或者过分割,提出基于类间蚂蚁竞争模型的显著图像分割算法。首先根据线性迭代聚类超像素分割算法(simple linear iterative clustering,SLIC)对图像进行预处理,在保留原始图像信息的前提下,将图像分割成各个区域,这样不仅可以提高分割精度得到理想的分割结果,还可以缩短运算时间。同时为了弥补单类蚂蚁分割易造成的欠分割或者过分割,引入两类蚂蚁,每一类蚂蚁寻找各自目标(前景/背景),不同类别的蚂蚁之间进行信息互补与竞争,使得分割结果更加准确。根据种群竞争思想,设定两类蚂蚁,每类蚂蚁设定食物目标不同,从而相互竞争,“优胜劣汰”,最终找到各自的食物,根据两类蚂蚁分泌的信息素竞争得到最终的结果。实验结果表明,该算法运行快速,分割结果更加精确。

Since the single ant colony algorithm is easy to cause under-segmentation or over-segmentation,we propose a saliency image segmentation algorithm based on ants competition model.Firstly,the simple linear iterative clustering(SLIC)is used to preprocess the image.Under the premise of retaining the original image information,the image is divided into various regions,which can not only improve the segmentation accuracy to obtain ideal segmentation results,but also shorten the operation time.At the same time,in order to make up for the under-segmentation or over-segmentation easily caused by the segmentation of a single class of ants,two types of ants are introduced.Each type of ants looks for its own target(foreground/background),and different types of ants complement and compete with each other for information,making the segmentation result more accurate.According to the idea of population competition,two kinds of ants are set.Each group of ants sets a different food target and competes with each other,“surviving of the fittest”to find their own food and compete for the final result based on pheromones secreted by the two groups of ants.Experiment shows that the proposed algorithm runs fast and segments accurately.