本文介绍了连通组件标记法，以8连接性为例，解释了如何从左到右、从上到下扫描图像，通过检查像素邻居来分配和合并标签，实现前景物体的识别和分组。
以 black-white and 8-connectivity 为例，0代表背景，1代表前景物体。该算法包括 two scans, from left to right and from top to down.
The first scan: examine the four neighbors(斜左上角,正上方,斜右上角和左方)
i. If none of the pixel value is 1, then assign this pixel a new label;
ii. If only one of this four pixels is 1, then assign its label to the current pixel;
iii. Then if more than 1 pixels are 1, then assign the minimal label to the current pixel, and make a note of equivalences for these pixels.
The second scan: merge the labels with the equivalent notes.
示例:
1. 源数据：
最低0.47元/天 解锁文章
扫一扫
2324
被折叠的 条评论
为什么被折叠?
到【灌水乐园】发言
