I can rule out that the pixels hold different pixel data...
If you make an image that is 100% white at 100x100 it will be the same size and an image with a different colour for each pixel at 100x100
even if you make an image with no colour at all.
Source: Adobe Illustrator cs4 Tutorial (from linda.com)
cant remember which.
This is only true if there is no form of image compression, as I stated above. Think of it this way, with three sheets of paper:
(1) The first sheet has the letter "A" printed all over it.
(2) The second sheet has English text printed over it.
(3) The third sheet has random English alphabet characters printed all over it.
Using an uncompressed TIFF algorithm to write that text, it would simply write every character as-is. Each sheet would take the same amount of data to store it.
Now let's say you used an algorithm that compressed based upon patterns. It sees the, say, 1000 letter "A" all in a row and can simply store this data as "1000xA" -- very short. It can recognize patterns in the English text in the second example and save every word "the" as a special character, thus helping to shrink the size of the text. But in (3), there is no way it can compress it, so saving that page would take up the most space. In this sense, every "image" (page of text) even though it's the same size will take up a different amount of space because it is compressed based on its content.
The same is true for pictures. If you have a quarter of your picture as over-saturated sky (so it's all white), then it is very easy for any even rudimentary compression algorithm to simply write that whole block as a giant rectangle that's just filled with "white" (255, 255, 255) if RGB 8-bit color as opposed to writing out the value of every pixel. And this is just a "lossless" compression.
In "lossy" compression schemes, the algorithm can cheat. Say there's a little bit of a tree sticking up into that white sky. A lossy algorithm may say that tree is too small to worry about and just ignore it, writing over it with the white sky to save room. Alternatively, it may decide to save a little bit of it as just a single-color straight line. This will take more information since it has to remember that, but it will take less than recording every single pixel value as it was originally. Hopefully you get the idea.
So again, it all depends on the compression algorithm, if there is one employed.
