[atishay811]

This is amazing to see. We should use this for image optimization. When we compress channels, we should compress the blue channel to like 30% while keeping others at fairly large 80% and it might appear better than a 60% compressed image.

[MaxBarraclough]

As triclops200 says, lossy image compression algorithms have long taken advantage of this. You might be interested in this page, Your Eyes Suck at Blue, which shows an image with the blue channel increasingly compressed:

https://gamesx.com/misctech/visual.htm

edit See also this counterpoint: https://news.ycombinator.com/item?id=573593

[triclops200]

We already do and have done historically. The most obvious example I know of is that 8bit color uses 3 bits for green and red and 2 for blue.

[_0ffh]

For some time many graphics cards had a 16 bit "hi-color" mode with 5:6:5 bits for RGB (SVGA, etc.). Most graphics card modes that used only 8 bits per pixel used that value as an index into an 256x(8x3) bit color palette (MCGA, VGA, etc.).

[nimish]

That's chroma subsampling, most commonly 4:4:4 4:2:2 and 4:2:0 where this principle is used

[eurasiantiger]

I guess we can call it lossy compression ;)

[trollbridge]

That’s exactly why images tend to be compressed in a YUV colour space instead of an RGB one.