黑素瘤图像毛发遮挡信息的非监督修复

Unsupervised repair of hair-occluded information for skin melanoma image

毛发遮挡信息的修复是正确分割和分析皮肤恶性黑素瘤图像的关键。针对皮肤镜黑素瘤图像,提出了一种毛发遮挡信息的非监督修复算法。该算法包含3个步骤:首先,运用形态学闭top-hat算子对毛发进行增强,并采用统计方法对图像进行二值化;然后,根据连通区域的延伸性特征提取出毛发;第三,运用基于偏微分方程的图像修复技术对毛发遮挡部位的信息进行修复。采用形态学增强方法,能同时增强强弱不同的毛发目标,所定义的延伸性特征函数能够描述条带状连通区域的延伸状态,是提取毛发的有效测度,尤其是将毛发提取与图像修复技术结合起来,解决了毛发的自动去除问题。实验结果表明:所提出的算法能够自动去除毛发,所修复的图像满足医学专家的判定要求,同时算法具有准确、鲁棒性强的特点。

The hair-occluded information repair is one of key problems for the precise segmentation and analysis of the skin malignant melanoma image with hairs. Aimed at skin melanoma images, an unsupervised repair algorithm for the hair-occluded information is proposed in this paper. This algorithm includes three steps： firstly, the melanoma image with hairs is enhanced by morphologic closing-based top-hat operator and then segmented through statistic threshold ; secondly, the hairs are extracted based on the elongate of connected region ; thirdly, the hair-occluded information is repaired by the image repair technology based on partial differential equation （PDE）. As a matter of fact, with the morphologic closing-based top-hat operator both strong and weak hair can be simultaneously enhanced, and the elongate state of band-like connected region can be correctly described by the elongate function proposed in this paper so as to measure the hair effectively. Therefore, the unsupervised repair problem of the hair-occluded information can be very well resolved through combining the hair extracting with the PDE-based image repair technology. The experiment results show that the repaired images can satisfy the requirement of medical diagnosis by the proposed algorithm and meanwhile some properties of well veracity, trustiness and robustness are manifested.