一种基于视觉显著模型的PET图像快速分割算法

A Fast Segmentation Algorithm of PET Images Based on Visual Saliency Model

由于PET图像质量较低且灰度相近导致常用的分割算法不能兼顾分割效果与分割效率,提出了一种基于视觉显著模型的PET图像分割算法。首先使用优化的Itti视觉显著模型替代人工操作对PET图像进行快速识别,并将获得的显著图进行预处理,初始化前景区与背景区的高斯混合模型,最后使用优化的GrabCut算法对PET显著图进行分割获得结果。与其他两种算法相比实验结果表明,该算法在操作简易性、算法执行效率、结果准确性具有一定优势,分割时间提升超过30%,分割精度提升超过21%。

For PET image has characteristics of low quality and similar gray led to the current main image segmentation algorithm unable to effectively take into account the effect and the efficiency of segmentation, this paper presents the segmentation algorithm which is based on visual saliency model of PET images. First and foremost, we use optimized Itti visual saliency model to identify PET images so as to replace manual operation and improve the speed of recognition; Next, pre-process the saliency map, and initialize the Gaussian mixture model of foreground area and background area; Finally, use optimized GrabCut algorithm to segment the PET saliency map and get PET image segmentation results. Compared with the other two algorithms, experimental results show that the proposed algorithm has some advantages in the simplicity of operation, the efficiency of the algorithm and the accuracy of the results, the segmentation time is increased by over 30% , and the segmentation accuracy is increased by more than 21%.