Current Algorithm seems to be based around Linear resize.

However when you look at other CBR's such as CDisplayex they have Lanczos implementation which is clearly superior interpolation at all zoom levels.



Suggestion 1: Add a better Scaling Algorithm.
It doesn't have to be Lanczos. It can be Spline 36/64 etc.

However there is clearly a lack of sharpness with the current algorithm compared to other implementations using the far superior interpolation.

As far as pc resources are concerned, even Laptops from 2013 have no issue scaling 4K native with Lanczos.
It really shouldn't be resource intensive at all to implement and if it is, that's what the pageflow performance tab is there for.



Suggestion 2: Auto color correction
A lot of .cbz's .rar etc are based on personal scans etc. However because most people don't know how to use a scanner, what often happens is the black point gets lifted. Scenes that should have pure blacks now have black levels starting at 50/255 which just lowers contrast for the entire image.

Yes there is an image adjustment tab. However, it has to be manually tuned per image.
Again, looking at CDisplayex's implementation. It is auto set per image. Which is a far, far better system.

What we're currently using is the stock image scaling algorithm provided by Qt, which to my best knowledge is a linear resize with bilinear smoothing.

We've looked into adding better scaling algorithms a few times during the last years and we've experimented with several approaches, including moving image scaling to the graphics card completely. Some of the reasons those attempts never made it into a feature were:

- Difficulty to integrate with our current page rendering process
- Difficulty to integrate custom scaling in stock Qt widgets
- Adding too much latency (GUI feels slow, even on modern CPU)
- Not portable enough

That does not mean it can't be done, but it is one of the more complicated features on our agenda, so we really can't say when we will work on it or when it might be finished.

Image enhancement/Auto color correction is a different issue. I've experimented with this and my implementation was fast enough to make it into the app, even on older CPU and without hardware acceleration. The reason why we finally opted not to integrate this feature for now was that the results were not up to standard.
I compared the output image of my simple autocontrast algorithm with the output of similar features in graphics programs and even after rechecking the math twice my image still came out different and the quality was worse.