形状描述符
我有一个图像，它是一个二元蒙版，我想表征蒙版中黑色区域的轮廓，我发现了几个功能来检测两个区域之间的边界，但只是作为一个新图像，但我想要的是描述这个边框，不是一个新的图像。 Java 高级成像或其他库中是否有任何功能可以应用于掩模以检索黑色区域的几何形状的表示？ 表示可以是形状的矢量化、多边形近似、链码……我所找到的只是该问题的理论解决方案。
I have an image which is a binary mask, I want to characterize the contours of the black zones in the mask, I have found several functionalities to detect the borders between both zones but only as a new image but what I want is the description of this border, not a new image. Is there any functionality in Java advance imaging or in another library which can be applied on the mask to retrieve a representation of the geometrical shape of black zones? The representation can be either a vectorization of the shape, a polygonal approximation, chain code... All I have found about is theoretical solutions for the problem.
如果你对这篇内容有疑问，欢迎到本站社区发帖提问 参与讨论，获取更多帮助，或者扫码二维码加入 Web 技术交流群。
绑定邮箱获取回复消息
由于您还没有绑定你的真实邮箱，如果其他用户或者作者回复了您的评论，将不能在第一时间通知您！
发布评论
评论（2）
您是否愿意编写一个遍历边界并收集边界像素列表的例程？ 如果你有这样一个列表，它会让你形成你想要的描述符吗？
Are you comfortable writing a routine that walks the boundary and collects a list of the boundary pixels? If you had such a list, would it let you form the descriptor you have in mind?
您可以尝试基于直方图的描述符，例如 SIFT、SURF 等。在 OpenCV 中，几个高级描述符是 已实现
如果你的描述符需要仿射协变（尺度和旋转不变），我强烈建议你尝试 SIFT。 本地二进制模式也是一个有效的选择。
You may try histogram based descriptors such as SIFT, SURF etc. In OpenCV, couple of advanced level descriptors are implemented
If your descriptors requires to be affine-covariant(scale and rotation invariant), I strongly recommend you to try SIFT. Local binary pattern is also an efficient choice.