CODEWITHSUNDEEP
c++opencvimage-processing
Gilad
Gilad

Reputation: 6595

openCV RGB2Gray with custom weights

The color to grayscale algorithm is stated in the cvtColor() documentation. (search for RGB2GRAY).

Y = 0.299 R + 0.587 G + 0.114 B

I am using 

    cv::Mat i_test_base;
    cv::cvtColor(white_balance_image, i_test_base, CV_RGB2GRAY);
    cv::imshow("cvtColor", i_test_base);
    cv::waitKey(0);

I'm working with Macbeth chart and it can result too low signal level when detecting e.g. blue patch. Using just sum between channels or max of the RGB values would most likely work better.

I tried something like 

cv::transform(white_balance_image, i_test_base, cv::Matx13f(1, 1, 1));
cv::imshow("transform", i_test_base);
cv::waitKey(0);

but this doesn't work, I get a white image. do I need to scale or something like that?

Upvotes: 2

Views: 2290

Answers (1)

beaker
beaker

Reputation: 16810

In your cv::transform, make sure that the weights sum to 1 (or less) to avoid saturation.

If you use 

cv::transform(white_balance_image, i_test_base, cv::Matx13f(0.114, 0.587, 0.299));

on a BGR image you should get the same as BGR2GRAY. If you use (1/3.0, 1/3.0, 1/3.0) instead, you should get an average of the 3 channels. Adjust to your liking.

Upvotes: 2

Related Questions