Thread Rating:
  • 0 Vote(s) - 0 Average
  • 1
  • 2
  • 3
  • 4
  • 5
Best algorithm for resizing images?
#1
Photo 
I have some images saved, in the past I reduced the size of the images a lot to save disk space (for example, I reduced an image from 3840x2160 to 1280x720, that is, 200% reduction).

Now I need to go back to the original size, and well, that's a problem, because I kept these images for other people, and these people want the original image (they cannot suspect that the image quality has been reduced).

I know it is impossible to increase with the original quality, but what is the best algorithm to increase the size of images in Gimp to get as close to the original? To me, NoHalo looks the best, the others seem to blur the image.

There are some sites that promise to increase quality using AI for free, but what is the guarantee of privacy? I don't want private images stored or posted around. And it is very suspicious of a website doing this for free.
Reply
#2
Upscaling result always in lost of quality bat you can try Reshade : https://www.reshade.com
If you want save disk space, a better solution is changing the jpeg quality (to 95% /90%)
In gimp, you can change the quality when you export the image :
     
Reply
#3
Quote:..I have some images saved, in the past I reduced the size of the images a lot to save disk space (for example, I reduced an image from 3840x2160 to 1280x720, that is, 200% reduction..

That is one-third the size of the original, two-thirds of the image information has been thrown away, never to come back.

Quote:....these people want the original image (they cannot suspect that the image quality has been reduced...

This is going to depend on their memory and their eyesight Wink

Quote:... what is the best algorithm to increase the size of images in Gimp to get as close to the original? To me, NoHalo looks the best, the others seem to blur the image.

Whatever you do, there is interpolation to make the new pixels. It will make the image softer, some methods more than others.

The 'halo' algorithms come from Nicolas Robidoux's work see: http://www.imagemagick.org/Usage/filter/nicolas/ In the past I went through most of these (mainly for down scaling). Not much difference in normal use, NoHalo is about as good as any method.

If I was trying to save disk space by scaling an image down, the last image format I would use is jpeg. Not much difference between a 100% jpeg and a lossless webp. The other lossless formats png / tiff not far behind. However storage is not expensive these days, no real reason not to keep the originals.

edit: ..but if the 'owner' is expecting a jpeg not much you can do.
Reply
#4
(08-25-2020, 08:01 AM)rich2005 Wrote:
Quote:..I have some images saved, in the past I reduced the size of the images a lot to save disk space (for example, I reduced an image from 3840x2160 to 1280x720, that is, 200% reduction..

That is one-third the size of the original, two-thirds of the image information has been thrown away, never to come back.

Hmmm. 3840×2160=8294400 (a bit over 8Mpix), 1280×720=921600 (a bit under 1Mpx). So the result is 11.1111%
of the original, in other words close to 90% of the image has been thrown away.
Reply
#5
(08-25-2020, 09:33 AM)Ofnuts Wrote: Hmmm. 3840×2160=8294400 (a bit over 8Mpix), 1280×720=921600 (a bit under 1Mpx). So the result is 11.1111%
of the original, in other words close to 90% of the image has been thrown away.

Why so picky ? I was referring to the size of the image: width x height  

   
Reply


Forum Jump: