This forum is disabled, please visit https://forum.opencv.org
| 2015-11-08 20:01:23 -0600 | received badge | ● Taxonomist |
| 2012-08-04 01:31:24 -0600 | received badge | ● Student (source) |
| 2012-08-01 18:44:01 -0600 | asked a question | Behavior of "Not-a-Number" (NaN) values in remap In the cv::remap function, is the behavior specified when values in the "map" arrays are NaN (not-a-number)? Or is the behavior unspecified and having NaN's in the "map" arrays should simply not be done? The behavior I am seeing (v2.4.1) is pixels where the map value are NaN, the destination value corresponds to the "source image" value at coordinate (0, 0). This appears to be true regardless of the border mode I use. |
| 2012-07-14 14:20:41 -0600 | received badge | ● Notable Question (source) |
| 2012-07-13 20:37:58 -0600 | received badge | ● Editor (source) |
| 2012-07-13 20:37:08 -0600 | received badge | ● Popular Question (source) |
| 2012-07-12 17:25:06 -0600 | asked a question | filter2D and sepFilter2D, specifying the value for a constant border When calling filter2D and sepFilter2D, I want to use a constant border, which I do by providing BORDER_CONSTANT for the "borderType" parameter as specified in the documentation (http://docs.opencv.org/modules/imgproc/doc/filtering.html?highlight=gaussianblur#filter2d). However, I do not see how I can specify WHAT I want that constant value to be. How do I achieve this? (edit: I did find that the default constant border value seems to be 0.) (edit in reply to Alexander, I do not see how to reply directly to answers: If I am not mistaken, copyMakeBorder will allocate a new image and copy the input image to a subregion of that new image. That seems a little excessive just to get access to the the ability to have non-zero constant border interpolation methods. Is there any other way to change the constant border value that does not require such memory allocations and copies?) |