基于Mean Shift和遗传算法的皮肤镜图像分割(英文)

Segmentation of dermoscopy images based on mean shift and genetic algorithm

皮损目标的精确分割是实现皮肤镜肿瘤图像自动分析的关键。提出了一种皮肤镜图像分割方法。首先采用一种优化的对比度增强方法进行预处理; 然后采用具有较好边界结构保持性的Mean Shift 算法对图像进行粗分割; 根据皮损区域和背景皮肤区域的信息,提出一种子区域合并的目标函数,并采用遗传算法对合并结果进行优化,获得满意的分割结果。实验结果表明,本文方法实现了皮肤肿瘤图像的有效分割,并在准确性上具有更优良的性能。

Accurate segmentation of the skin lesion plays an important role in automatic analysis of the dermoscopy image. A segmentation method for the skin tumor image is presented. Firstly, the images are preprocessed by an optimal contrast enhancement method. The Mean Shift algorithm is then adopted to obtain the coarse segmentation result as its good boundary retention performance. According to the information of the lesion and background skin regions, we incorporates genetic algorithm into the sub-regions merging process. Finally, statistical measures are introduced to evaluate the performance of segmentation. Experimental results show that our proposed method can segment the lesions from the surrounding skin effectively and yield competitive accuracy compared with other four automatic segmentation methods, including Otsu＇s threshold, k-means, fuzzy c-means （FCM） and statistical region merging （SRM）.