Write a program to change brightness and change the contrast of the color image (keep the average brightness of the image). I don't understand how to keep the average brightness of the image. thank you!!!

"Write a program to change brightness and change the contrast of the color image (keep the average brightness of the image)."

I don't understand how to keep the average brightness of the image. thank you!!!

Answer 1

[Walter Roberson]

The way to keep the average brightness of an image is:

1. calculate the average brightness of the original image; Call this BRorig
2. make whatever local changes to color and contrast;
3. calculate the average brightness of the revised image; Call this BRnew
4. multiply the revised image by the ratio BRorig/BRnew

[John D'Errico]

With the caveat that this solution may easily fail completely.

A simple linear (multiplicative) remapping of the pixels can cause some or many of the pixel intensities to block up beyond their maximum. Suppose the pixels are stored as uint8 integers, thus 0-255. Anything at or near white, thus [255,255,255], will overflow. So you won't actually achieve the required brightness change. It can also change the colors pixels with some components near the maximum intensity. For example, a 15% multiplicative rescaling of a pixel that was at [222,0,255], will remap to [255,0,255]. So only the red component of that pixel ended up changing, thus a color shift results. Expect to see arbitrary color changes for any pixels that had one or more components at or near their maximum.

I looked online and saw that much of the advice given for brightness remapping was to use an additive shift. Similarly, simple additive remapping of the pixels can also cause overflows or underflows. Again, that can result in arbitrary color shifts, regions that block up at white or black, etc.

You will be somewhat better off if you remap the image into a color space like HSV, L*a*b*, etc. Now you can do brightness remapping by modifying only one of the image planes. Then after the remapping, you convert back into an RGB color space. Even here there may be gamut issues, with some pixels blocking up in corners of the RGB space.

The point is, such simple linear (thus additive or multiplicative) brightness changes done purely in RGB can cause problems. You may not achieve the desired brightness shift, and you may observe hue shifts. Where a large shift is necessary, expect the result to look very bad, but if the shift is a small one you may be ok.

[Hoang Nguyen]

Walter Roberson, Sorry guy, Can you help me? how to calculate the average brightness of the original image, I don't understand "Call this BRorig" because I don't find to calculate the average brightness

[Walter Roberson]

BRorig = mean2( rgb2gray( YourColorImage ) );

[Hoang Nguyen]

John D'Errico, My teacher said: 'Used to HSI, but I don't understand' My English not good Can you contact me with facebook: "[e-mail]"? Thank you!! =))

[Walter Roberson]

You could also calculate the total brightness of the original image (instead of the average) and the total brightness of the modified image, and then pick random locations to add increments to, taking into account that brightness = 0.2989 * R + 0.5870 * G + 0.1140 * B

Answer 2

[Image Analyst]

I'd use a different approach. You can use rgb2hsv to convert to HSV color space. Now, it really depends on what your teacher means by contrast. One way is to increase the brightness contrast and to do that you'd subtract the mean from the V channel, multiply by some value, add back in the original mean, and then transform back:

hsvImage = rgb2hsv(rgbImage);
hChannel = hsvImage(:, :, 1);
sChannel = hsvImage(:, :, 2);
vChannel = hsvImage(:, :, 3);
meanV = mean2(vChannel);
newV = meanV + 1.5 * (vChannel - meanV); % Increase contrast by factor of 1.5
newHSVImage = cat(3, hChannel, sChannel, newV);
newRGBImage = hsv2rgb(newHSVImage);

The other thing they might mean is to increase the saturation so the colors look more vivid. In that case, you'd just multiply the S channel by some factor.

hsvImage = rgb2hsv(rgbImage);
hChannel = hsvImage(:, :, 1);
sChannel = hsvImage(:, :, 2);
vChannel = hsvImage(:, :, 3);
newS = 1.5 * sChannel; % Increase contrast by factor of 1.5
newHSVImage = cat(3, hChannel, newS, vChannel);
newRGBImage = hsv2rgb(newHSVImage);

Of course, clipping may occur if you go back to uint8 (as John said).

Or you could change both the V and S channels.

[Hoang Nguyen]

Thank you for your answer :D :D

[Akib Rahman]

Image Analyst, when I write same code but it gives me some error :( I tried so many ways, but it also remains wrong. Would you help me? ERROR:

• Error using rgb2hsv>parseInputs (line 95)MAP must be a Mx3 array.
• Error in rgb2hsv (line 36)[r, g, b, isColorMap, isEmptyInput, isThreeChannel] = parseInputs(varargin{:});
• Error in brightness_contrast_change (line 1)hsvImage = rgb2hsv('image19.jpg');

[Image Analyst]

I don't think it takes a string. It needs to take an image so use imread() to read 'image19.jpg' into an image:

rgbImage = imread('image19.jpg');
hsvImage = rgb2hsv(rgbImage);

By the way, don't use JPG format for image analysis - it introduces artifacts that will affect your measurements.

[Akib Rahman]

Thanks, Image Analyst :) It works.

Answer 3

[Rodrigo Souza]

Hi all,

I have put together three scripts/functions for calculating (mean and SD), matching or normalizing luminance of colored images (using HSV and CIE Lab color spaces).

They may be specially useful for pupillometry measures of infant research.

The scripts and their descriptions are available at my OSF page: https://osf.io/auzjy/

Hope it helps.