After I used Photozoom to enhance this photo, I found it really nice for a wallpaper, the color is comfortable. Of course if she could put on a pantsu, it would be more suitable for the wallpaper......

"photozoom" and "enhance" are mutually exclusive.

PhotoZoom(v3.1) is a software with patent resize arithmetic called S-Spline, which has the best quality on resizing a low res photo to a higher res. I have used this software for years, sometimes it really helpful when you find a good photo only with 1024x768 but you want to print it out in high quality.

Upscaling something never results in "high quality" only higher resolution.

Ponnkun said:
Upscaling something never results in "high quality" only higher resolution.

That's exactly what this software can do, it's the selling point. Compare to other upscaling arithmetics, you can see the advantages of S-Spline(S-Spline XL/MAX), not perfect, but really high quality.

Drich007 said:
That's exactly what this software can do, it's the selling point. Compare to other upscaling arithmetics, you can see the advantages of S-Spline(S-Spline XL/MAX), not perfect, but really high quality.

you mean a better quality than just plain upscaling

That's true, you can try "S-Spline MAX" upscaling arithmetic in PhotoZoom v3.1 to upscale a 1024x768 photo to 4000x3000, then compare with what was done in ACDSee, then you could find that it is really a enhancement, not a simply upscaling.

it seems re-calculate lines then the result hasn't aliasing, works on vectors very well.
blurness and surface noises are the point to detect upscaling, but it's head aching when someone start posting upscaled with them :|

midzki said:
it seems re-calculate lines then the result hasn't aliasing, works on vectors very well.
blurness and surface noises are the point to detect upscaling, but it's head aching when someone start posting upscaled with them :|

Turn the "Film Grain" to 0, you will see no noise.
I upscaled this photo to 4000x3000 using S-Spline MAX, then used ACDSee to resize the upscaled one to 1920x1440, which has better sharpness and detail than the 1600x1200 version in Konachan...... Surely I could not tell which was the raw or upscaled one according to the quality...

Drich007 said:
Turn the "Film Grain" to 0, you will see no noise.
I upscaled this photo to 4000x3000 using S-Spline MAX, then used ACDSee to resize the upscaled one to 1920x1440, which has better sharpness and detail than the 1600x1200 version in Konachan...... Surely I could not tell which was the raw or upscaled one according to the quality...

"Better Details"
How could you increase the details if they don't exist?

I think different people has different definitions of "enhancing", but at least this will come in handy when you want a large printout or wallpaper and there is no higher native res available.

In fact, greycstoration can also do resampling, wonder if anybody has tried it out yet?

There should be a ultimate goal for image upscaling from the point of view of digital data, maybe it's loseless upscaling? like a vector (of cause, I don't mean the normal vector).

But, humans aren't machines, the best reproduction of an image doesn't mean the best feeling for people. Thus, some "tricks" in process will works

Drich007 said:
Turn the "Film Grain" to 0, you will see no noise.
I upscaled this photo to 4000x3000 using S-Spline MAX, then used ACDSee to resize the upscaled one to 1920x1440, which has better sharpness and detail than the 1600x1200 version in Konachan...... Surely I could not tell which was the raw or upscaled one according to the quality...

could you upload them here with hold temp tag? (hold posts can be deleted by a poster anytime even past 24 hours)
I just tested with a trial version which can't export without a watermark :|

kiowa said:
In fact, greycstoration can also do resampling, wonder if anybody has tried it out yet?

too slow and can't expect its results because there is no GUI version. few results I could get were similar to bicubic+greycstoration.

fireattack said:
"Better Details"
How could you increase the details if they don't exist?

I know your meaning, I only want to say that details are more clear. And the sharpness is natural, not incondite as that sharpened in PS.

could you upload them here with hold temp tag? (hold posts can be deleted by a poster anytime even past 24 hours)
I just tested with a trial version which can't export without a watermark :|

Here you are: post #155984

Please don't post links to cracked software - Radioactive

thank you (though I meant hold + temp ^^;). I'll compare carefully later

Drich007 said:
I know your meaning, I only want to say that details are more clear. And the sharpness is natural, not incondite as that sharpened in PS.

Here you are: post #155984

Go download a cracked version with mutilanguage: http://u.115.com/file/f03835b16 (a link from china, maybe slow. Click "网通1下载" )

I checked it quickly, the lines are a bit strange (too shaper and unnatural), but that's based on the fact I did know it's not original. S-Spline is indeed a great algorithm, actually I used it for video upscaling.

fireattack said:
I checked it quickly, the lines are a bit strange (too shaper and unnatural)

Try turn off the unsharp masking ( or try the S-Spline XL).
Or if you want to upscale a photo to 4000x3000, you can even upscale it to 8000x6000, then use ACDSee to downscale it to 4000x3000, maybe the lines appear better.

It was more cleaned because jpeg_artifacts were declined (that means this level of details will be lost)
It's not hard to tell which lines are curving "naturally", but if only when I have an original to compare.

Yes, I know that BenVista also has a software to denoise the jpeg_artifacts, maybe it's contained in S-Spline MAX. Using S-Spline or S-Spline XL will keep more details, but show more aliasings on lines.

If you want to keep more details with S-Spline MAX, you should do this:

Or if you want to upscale a photo to 4000x3000, you can even upscale it to 8000x6000, then use ACDSee to downscale it to 4000x3000

Then you can find jpeg_artifacts remain there.