基于选择性注意机制的无边界主动轮廓图像分割算法

针对现有的主动轮廓算法多为人为给定初始轮廓线的问题,提出一种基于选择性注意机制的无边界主动轮廓图像分割算法。该算法首先利用数学形态学算子对图像进行平滑预处理,然后将选择性注意机制引入到基于区域的无边界主动轮廓模型中,对单目标和多目标图像分别采用基于显著图的方法以及基于选择注意与小波变换相结合的方法进行掩膜初始化,最后应用水平集方法进行图像分割。结果表明,该算法不仅可以减少迭代次数,当图像中存在多个目标时还可以得到更精确的分割结果,有效地提高了主动轮廓算法的效率。

Active contour without edges for image segmentation based on selective attention

Given the fact that in the majority of current active contour algorithms the initial contour has to be given manually, an improved active contour without edges algorithm for image segmentation is put forward based on selective attention. Firstly, a mathematical morphology operator is used for smoothing preprocessing, then a selective attention mechanism is incorporated into the region-based active contour model without edges. Depending on whether it is for a single target image or a multiple-target image the saliency map based algorithm or the algorithm utilizing selective attention with wavelet transform is used to initialize mask, respectively. Lastly, the level set is applied for image segmentation. Experimental results show that the proposed algorithm not only reduces the overall number of iterations, but also obtains more accurate segmentation results for multiple-target image, improving the efficiency of active contour algorithm significantly.