get RGB data from image

I was working on some code for a small assignment and can't seem to figure out why I can't get my bright vector to come out as [251,251,251] using this attached picture. The RGB values for the white section of the image are all 251.

picture = imread('ColorWheel.png')    

[a,b,c] = size(picture);

bright = [0,0,0];
    
    for y = 1:b
        
        for x = 1:a
            
            if picture(x,y,1) >= bright(1) && picture(x,y,2) >= bright(2) && picture(x,y,3) >= bright(3)
                bright(1) = picture(x,y,1);
                bright(2) = picture(x,y,2);
                bright(3) = picture(x,y,3);
            end
        end

Answer 1

[Walter Roberson]

Your search encounters a pixel with components 5 6 255 . Your test for assigning a new brightness requires that **all* of the r g b components of the candidate be at least as big as the current bright . The pixels with value [251 251 251] fail on 251 >= 255 .

If the purpose of the assignment is to find the location with maximum brightness, then you need to take into account that brightness does not vary equally with the different color components. The green component has the most influence on brightness, followed by the red component, and the blue component has relatively little impact on brightness.

[Sorin Peña]

I changed it to accomodate this, but I still am encountering issues. The code is now adjusting for each individual color value, but if I try to index a point ex. (220,60,1) I get a uint8 value of 1. I don't understand quite how uint8 works, but I thought if I converted that to a double I would get a value of 251.

[Walter Roberson]

If you work one component at a time then you risk ending up with a triple that does not exist in the image.

[Image Analyst]

Do you want the brightest overall? Or the brightest for each color channel independently? For example (251, 251, 255 might be the brightest overall even though there are other pixels, like (255, 0, 0) and (0, 255, 0) that are brighter in one separate color channel.

[Walter Roberson]

uint8 tells you about data representation. The 8 tells you that each value in the array is represented by 8 bits, and the int tells you that an integer encoding is being used, and the u tells you that unsigned values are being used. uint8 says that each byte of memory is to be interpreted as an value that is an integer in the range 0 to 255 and that no negative values are permitted and no values greater than 255 are permitted.

double also tells you about data representation, though not as directly. double is short for "double precision", which is a binary floating point representation that has approximately twice the precision as "single" precision binary floating point representation does. MATLAB documents that it uses IEEE 754 Single and Double Precision for its "single" and "double" . IEEE 754 is a particular international standard number. It uses 1 bit a separate sign, 11 bits for exponent, 52 bits for mantissa, and 1 "hidden" bit (64 total bits) It works like interpreting the mantissa as a 52 bit unsigned integer, adding 2^53 (putting a binary "1" before the number), and dividing the whole thing by 2^(53+exponent-fixed_offset) -- so it is based upon integer multiples of a power of 2 (including negative powers of 2). For example, 83/64 could be represented exactly... but 1/10 cannot be represented exactly.

[Sorin Peña]

They said brightest, but they actually mean whitest. How could I index the point and actually return the color values? When I index (220,60,:) for example, I get 1, 114, and 62. I also have to use a for loop, and can't use any functions such as max.

[Walter Roberson]

Those 1, 114, and 62 are the color values. That translates as red component being 1/255 times the maximum possible red, and green component being 114/255 times the maximum possible green, and blue component being 62/255 times the maximum possible blue.

Note that if what they actually mean is "whitest", then you would measure white content not by high value but rather by how equal the components are. As in

temp = sort([red(:), green(:), blue(:)], 2);
offwhiteness = reshape( sqrt((temp(:,2) - temp(:,1)).^2 + (temp(:,3) - temp(:,2)).^2), size(red));

and you would target offwhiteness of 0. A location that is [183, 183, 183] is exactly as "white" as a location that is [19, 19, 19], because in both cases there is an equal mixture of frequencies. The difference between them is brightness, which you say they are not actually looking for.

I would suggest to you that what they are really looking for is indeed brightness, but that you have misunderstood how brightness is calculated.

A pixel which is [128, 0, 0] is brighter than a pixel which is [0, 0, 255]. The first of those would work out as about 15% of maximum brightness, whereas the second of them would work out as about 11% of maximum brightness.

[97 0 0] is about as bright as [0 50 0] which is about as bright as [0 0 255]

[Sorin Peña]

I understand now. Thank you all, my code is working well.