Inconsistent and unexpected behavior of cv.multiply() function with color images across OpenCV versions (4.5 to 4.8 vs. 4.9) about opencv HOT 3 OPEN

Maybe duplicated #25165

from opencv.

Umm, I read again. This behavious has been changed after 4.9.0, but I think OpenCV 4.9.0 behavious is correct

( More precisely, the case where src2 is Scalar is not described in the multiply function... )

1. When using an integer as the multiplier, versions 4.5 to 4.8 only the blue channel is correctly multiplied by 2.0, while the green and red channels' pixel values become zero.
2. When using a one-dimensional NumPy array (np.array([2.0])) as the multiplier, versions 4.5 to 4.8 correctly multiply all channels (BGR) of the image by the given value. However, in OpenCV 4.9, only the blue channel is correctly multiplied, while the green and red channels' pixel values are set to zero.

Please could you see atirhmertric function behavious at https://docs.opencv.org/4.9.0/d2/de8/group__core__array.html#ga10ac1bfb180e2cfda1701d06c24fdbd6 ?

• Precondition : src1(I) = [50, 50, 50]
• Try to execute : dst = multiply( src1 , src2).
• If src2 = 2 then dst = [50x2, 50x2, 50x2] = [100, 100, 100]
• If src2 = [2,3,4] then dst = [50x2, 50x3, 50x4] = [100, 150, 200]
• If src2 = [2] then dst = [50x2, 50x0, 50x0] = [100, 0, 0]

From https://docs.opencv.org/4.9.0/d2/de8/group__core__array.html#ga979d898a58d7f61c53003e162e7ad89f

Note
Saturation is not applied when the output array has the depth CV_32S. You may even get result of an incorrect sign in the case of overflow.
(Python) Be careful to difference behaviour between src1/src2 are single number and they are tuple/array. multiply(src,X) means multiply(src,(X,X,X,X)). multiply(src,(X,)) means multiply(src,(X,0,0,0)).

And This behavious is same as C++

#include <iostream>
#include <opencv2/core.hpp>

int main(void)
{
  cv::Mat gray_img(400,400,CV_8UC3, cv::Scalar::all(50));
  cv::Mat tmp_img_int;
  cv::multiply(gray_img, 2, tmp_img_int);
  std::cout << "INT = " << cv::format(tmp_img_int(cv::Rect(0,0,2,2)), cv::Formatter::FMT_NUMPY) << std::endl;

  cv::Mat tmp_img_scalar;
  cv::multiply(gray_img, cv::Scalar(2), tmp_img_scalar);
  std::cout << "SCALAR = " << cv::format(tmp_img_scalar(cv::Rect(0,0,2,2)), cv::Formatter::FMT_NUMPY) << std::endl;
  return 0;
}

Result

$ make && ./a.out
g++ main.cpp -o a.out \
        -I/usr/local/include/opencv4 \
        -lopencv_core
INT = array([[[100, 100, 100], [100, 100, 100]],
       [[100, 100, 100], [100, 100, 100]]], dtype='uint8')
SCALAR = array([[[100,   0,   0], [100,   0,   0]],
       [[100,   0,   0], [100,   0,   0]]], dtype='uint8')

from opencv.

@cocositu If OK, please could you close this issue ?

from opencv.