基于形态学重建的粘连物体分割

提出一种基于形态学重建（Morphological Reconstruction）的图像分割方法。该方法先对待分割图像进行预处理，使边界点具有局部极大的灰度值；然后利用灰度形态学重建提取穹顶（Dome），并根据其特性利用阚值对穹顶进行二值化获得候选边界点集；再利用二值形态学重建确定候选边界点集中的边界点，得到分割边界。实验结果表明，本分割方法所得边界连续性好、假边界少；该方法受噪声和对象内部灰度变化的影响较小，适合用于分割含有粘连对象的图像。

Segmenting Conglutinate Objects Based on Morphological Reconstruction

In this paper, a new segmenting approach to conglutinated object images is proposed based on morphological reconstruction. The approach first ascertains the foreground and the background in an image using the Ostu threshold method, and then processes the image to assure that the gray values of boundary pixels in the image are greater than those of the pixels nearby. Using grayscale morphological reconstruction, the approach extracts h-domes from the rows in the preprocessed image along different directions, and then binaries the h-domes with a threshold equal to h, getting boundary pixel candidates. The candidates consist of the boundary pixels of the objects and some pixels inside the ob- jects that are also local maxima in some rows and satisfy another constraint. Then the candidates and the background pixels are mapped into a binary image. In the binary image, the boundary pixels of the objects are connected to the background and other candidates that are inside the objects are not. According to that, the approach use binary mor- phological reconstruction to select boundary pixels from the candidates, getting the boundary for segmentation. Experi- ments show that the boundary got by the proposed segmenting approach has good continuity and fewer fake boundaries. The approach is less sensitive to noises and gray value's varying inside objects, and is suitable for segmenting images of conglutinated objects.