I'm working with the KITTI stereo dataset having 4 cameras. Two of them are grayscale and the other two are color. I need all four of them to be the same. Directly converting the color ones to grayscale does not give images of the same brightness. What's the correct way to convert them?