In GIMP, how can I identify and match the saturation of an image

Question

I have various album covers which are low resolution, but the colours come out really pretty.

I have higher resolution images of the same covers, but the colours are very different, much duller and less vibrant.

The first image is the original which I would like to imitate. The second is the larger resolution image I have acquired (scaled down for easier viewing). The last image on the bottom is my attempt at imitating the top image using the second image. Using GIMP, my settings were:

Colors → Saturation: 1.500
Colors → Brightness-Contrast → Brightness 35
Colors → Brightness-Contrast → Contrast 8

But as you can see, the colours and tones are just not quite the same. The skintone isn't as rosy-peachy and the colours just don't feel as warm as the original.

Answer 1

I think there is some major color profile confusion going on here.

When I download the low res version with the vibrant colors you like, I notice that its color profile is Adobe RGB. The hi res version is Untagged RGB but I assume it's sRGB.

Here are the two images side by side:

If I wrongly assign Adobe RGB to the hi res version, it gets more vibrant and starts to look like the lo res version:

And if I then convert the hi res version to sRGB and assign Adobe RGB once more, I actually get the exact same colors:

This might have been done deliberately by someone who believes it's a neat way to make an image more vibrant, or it has happened by mistake in some editing frenzy.

Either way, I think it's a bad idea to do like this. It looks unnatural, it's out of control and, as shown in another answer, it makes the image loose details in the colors.

The hi res version looks way better and more natural in my opinion. I would just keep the hi res version as it is and dump the overly saturated lo res version.

Answer 2

Looking at the pictures you posted, it seems to me that the ones you consider "duller and less vibrant" are actually higher quality images, not just in resolution, but in color information.

This is a detail taken from the "duller" image:

This is the same detail from the "more vibrant" one:

It seems obvious to me that a lot of color information was lost, or "squashed" in the preparation of the "vibrant" picture.

You can see exactly how much color information was lost if you compare channel histograms between the two images. This is the red channel (taken from the Levels dialog, but you can use Curves if you prefer) from the "duller" image:

And this is the same channel from the ruined ("more vibrant") picture:

If you compare the positions of the peaks, which I have highlighted with arrows, you will see how much information was squashed into 0 and into 255 in the preparation of the "more vibrant" picture.

If you want to achieve the same result (for whatever reason) you can use the Levels dialog and just trim the Input Levels (separately for each RGB channel, ignoring the composite "Values" one) until you achieve a similar histogram to the one in the "more vibrant" picture.

As for me, I would just use the better quality ("duller") pictures and if anything, I would apply a gamma adjustment "by eye" (in this case around 0.9)

Answer 3

You can apply the colors of the low-res image to the hi-res image quite easily, because our perception of the details is mostly based on luminosity and not on color or saturation(*).

• Open the high-res image

• File ➤ Open as layers the low-res image

• Scale it up to match the hi-res image. This will make it blurry. Here is for instance, your image, artificially blurred by scaling it down 4x (to 101*101px) and then re-scaling it up 4x:

  

• Align that image over the hi-res image. A good way is to set it to Difference blend mode and move it around until things get as dark as possible and edges as faint as possible:

  

• Then just change the blend mode mode to LCh color:

  

In that image the detail comes from the hi-res image (L channel: Luminosity) while the colors come from the C, h components of the low-res image.

But keep in mind that, with the same colors, a small picture can look more colored than a big picture because color changes are closer to each other.

(*) This is taken advantage of by the JPEG encoding ("chroma subsampling").