Abstract: In order to detect the faint and overlapped protein spots in protein gel images, a new approach based on the top-hat transform and shape characteristics is proposed. Top-hat transform is used to enhance faint protein spots region. A marker-controlled watershed algorithm is used to detect the initial outline of the protein spots. Then, shape markers are adaptively extracted by shape characteristics information, and the shape distance image is calculated according to these shape markers. The watershed algorithm is applied again on the shape distance image to separate the overlapped spots. The results of the experiments on different types of real gel images show that the algorithm has higher detection accuracy and better overlapped spots separation rate. It also obtains better detection results from low quality gel images.

Key words: gel images, top-hat transform, faint protein spots, overlapped protein spots, shape markers

摘要： 针对凝胶图像中蛋白质点检测存在弱蛋白质点漏检和重叠蛋白质点难分离的问题，提出了一种基于top-hat变换与形状特征的弱重叠蛋白质点检测算法。采用top-hat变换算法增强弱蛋白质点区域；采用标记控制分水岭法进行粗检测，提取蛋白质点的初始轮廓；根据蛋白质形状特征，自适应设定阈值提取蛋白质点形状标记，并利用提取的标记计算形状距离图；采用基于形状距离的分水岭方法，分离重叠蛋白质点。通过不同类型真实凝胶图像的蛋白质点检测实验，结果表明，该算法具有较高的检测精度和重叠蛋白质点分离率，而且对质量不好的凝胶图像也有较好的检测效果。

关键词: 凝胶图像, top-hat变换, 弱蛋白质点, 重叠蛋白质点, 形状标记