细化视差图？
谁能告诉我一种细化视差图的方法？我正在尝试生成人脸的视差图，但眼睛、鼻子、嘴唇等特征并不清晰。如何改进它以使其看起来更好？
Can anyone tell me of a method to refine disparity maps? I am trying to generate the disparity map of a face but the features like eyes, nose ,lips etc are not clear.How can I refine it to make it look better?
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）
看看
https://code.ros.org/trac /opencv/browser/trunk/opencv/samples/cpp/stereo_match.cpp
有一些 bm.state 参数（大约第 197 - 207 行）可以调整，我们将这些值连接到一些滑块，这样我们就可以在调整值的同时查看结果。您还可以尝试为“blocksize”使用一些不同的值。您还可以尝试在相机之间设置不同的距离，如果相机距离太远，近距离时您会得到很差/没有结果。
我假设您的代码与上面的示例类似。
Take a look at
https://code.ros.org/trac/opencv/browser/trunk/opencv/samples/cpp/stereo_match.cpp
there are some bm.state parameters (about line 197 - 207) that can be tweaked, we connected those values to some sliders so we could look at the result at the same time as we tweaked the values. you can also try with some different values for "blocksize". you can also try with some different distances between your cameras, if the cameras are too far apart, you will get poor / no results on close distances.
I assume that your code is similar to the example above.
使用stereoSGBM 我正在使用它，你可以使用轨迹栏来调整参数研究opencv的参考，然后根据它对图像的影响来操纵每个参数，例如：P1 P2使其更平滑等
Use stereoSGBM I am using it and you can use trackbars to tweak the parameters study the refernece of opencv and then manipulate each parameter based on the effect it has on your image eg:P1 P2 make it smoother etc