cv::COLOR_YUV2BGR and cv::COLOR_YUV2RGB are not handled in color.hpp dstChannels #18878
Description
System information (version)
- OpenCV => 4.5
- Operating System / Platform => Ubuntu 18.04
- Compiler => gcc 7.5
Detailed description
While doing some performance comparisons between cv::Mat and cv::UMat (OpenCL), I noticed that OpenCL was taking a lot longer (8x) when performing color conversions from YUV to BGR or RGB.
I compared my benchmarking code to the perf tests, and I was not populating the optional dstCn parameter in my calls to cv::cvtColor(). When I did pass that parameter, my performance greatly improved!
I think I tracked it down to the function dstChannels() in color.hpp
opencv/modules/imgproc/src/color.hpp
Line 109 in 049b50d
cv::COLOR_YUV2RGB and cv::COLOR_YUV2BGR codes are not in the switch-case list, so the default value of 0 is returned. This then gets passed to oclCvtColorYUV2BGR();
I don't fully understand why passing 0 as the value for dcn to the OpenCL kernel results in such poor performance, but seems like a simple fix.
Steps to reproduce
```.cpp
cv::UMat yuv = cv::UMat::zeros(1920, 1080, CV_8UC3);
cv::UMat rgb;
cv::cvtColor(yuv, rgb, cv::COLOR_YUV2RGB); /// This is slow :-(
cv::cvtColor(yuv, rgb, cv::COLOR_YUV2RGB, 3); /// This is fast!
```
I can whip up a complete test case/perf test if that's helpful.
Issue submission checklist
- I report the issue, it's not a question
- I checked the problem with documentation, FAQ, open issues,
answers.opencv.org, Stack Overflow, etc and have not found solution - I updated to latest OpenCV version and the issue is still there
- There is reproducer code and related data files: videos, images, onnx, etc