Skin melanoma cancer, particularly among non-Hispanic white women and men, has been one of the highest risks of spreading disease among all cancers. It should be treated earlier for effective treatment. Due to high costs of screening each patient by dermatologists, it is important to establish an automated method to determine the risk of melanoma for a patient by using image scan of their skin lesions that can provide accurate diagnosis. The major challenge is segmenting the skin lesion from the digital scan image. For segmenting the lesion, a novel algorithm based on skin texture is proposed in which a set of representative texture distributions is analysed from a non-illuminated image. The ridge in the skin image is labeled as either normal segment or lesion, based on the presence of sample texture distributions by calculating the texture distinctiveness metrics. In comparison with other bench-mark models the suggested algorithm has greater precision in segmentation about 95% accuracy.

中文翻译：

---

皮肤镜图像中皮肤病变的分割和分类

皮肤黑色素瘤癌症，尤其是在非西班牙裔白人女性和男性中，一直是所有癌症中传播疾病的最高风险之一。为有效治疗，应及早治疗。由于皮肤科医生对每位患者进行筛查的成本很高，因此重要的是建立一种自动化方法，通过使用可以提供准确诊断的皮肤病变图像扫描来确定患者患黑色素瘤的风险。主要挑战是从数字扫描图像中分割皮肤病变。为了分割病变，提出了一种基于皮肤纹理的新算法，其中从非照明图像中分析了一组具有代表性的纹理分布。皮肤图像中的脊被标记为正常段或病变，通过计算纹理独特性度量，基于样本纹理分布的存在。与其他基准模型相比，建议的算法在分割方面具有更高的精度，准确度约为 95%。