No effect from WebP's near-lossless option in ImageMagick convert

[alanorth]

ImageMagick version

7.1.1.5-1

Operating system, version and so on

Arch Linux

Description

The ImageMagick docs for WebP note that there should be a "near-lossless" option that maps to cwebp's -near_lossless. I tried to test that briefly and it doesn't seem to be working:

$ for q in 20 40 60 80 90; do magick convert data/10568-103447.pdf\[0\] -define webp:near-lossless=$q /tmp/10568-103447-near-lossless-q${q}.webp; done
$ ls -l /tmp/10568-103447-near-lossless-q*
-rw-r--r-- 1 aorth aorth 37470 Mar 29 14:23 /tmp/10568-103447-near-lossless-q20.webp
-rw-r--r-- 1 aorth aorth 37470 Mar 29 14:23 /tmp/10568-103447-near-lossless-q40.webp
-rw-r--r-- 1 aorth aorth 37470 Mar 29 14:23 /tmp/10568-103447-near-lossless-q60.webp
-rw-r--r-- 1 aorth aorth 37470 Mar 29 14:23 /tmp/10568-103447-near-lossless-q80.webp
-rw-r--r-- 1 aorth aorth 37470 Mar 29 14:23 /tmp/10568-103447-near-lossless-q90.webp

The file sizes are all the same.

I tried again with cwepb (from libwebp 1.3.0) and it works as expected:

$ magick convert data/10568-103447.pdf\[0\] -quality 100 img/10568-103447.pdf.jpg
$ for q in 20 40 60 80 90; do cwebp -near_lossless $q img/10568-103447.pdf.jpg -o /tmp/10568-103447-cwebp-near-lossless-q${q}.webp; done
$ ls -l /tmp/10568-103447-cwebp-near-lossless-q*
-rw-r--r-- 1 aorth aorth 115394 Mar 29 14:26 /tmp/10568-103447-cwebp-near-lossless-q20.webp
-rw-r--r-- 1 aorth aorth 122056 Mar 29 14:26 /tmp/10568-103447-cwebp-near-lossless-q40.webp
-rw-r--r-- 1 aorth aorth 118732 Mar 29 14:26 /tmp/10568-103447-cwebp-near-lossless-q60.webp
-rw-r--r-- 1 aorth aorth 140514 Mar 29 14:26 /tmp/10568-103447-cwebp-near-lossless-q80.webp
-rw-r--r-- 1 aorth aorth 140514 Mar 29 14:26 /tmp/10568-103447-cwebp-near-lossless-q90.webp

I see a related question about this from a few years ago, but I'm not sure which ImageMagick version shipped with support for this. Thanks!

[fmw42]

What version of webp is installed with Imagemagick

convert -list format | grep webp

     WEBP* WEBP      rw+   WebP Image Format (libwebp 1.3.0 [020F])

[alanorth]

Hey @fmw42, here's what is reported for me:

$ convert -list format | grep webp
     WEBP* WEBP      rw+   WebP Image Format (libwebp 1.3.0 [020F])

[fmw42]

Looks OK to me. But I am not an expert on WEBP. So wait for one of the IM developers to comment.

[urban-warrior]

Thanks for the problem report. We can reproduce it and will have a patch to fix it in the GIT main branch @ https://github.com/ImageMagick/ImageMagick later today. The patch will be available in the beta releases of ImageMagick @ https://imagemagick.org/archive/beta/ by sometime tomorrow.

Use -define webp:lossless=1 -quality 90, for example.

[alanorth]

Great, thank you! Regarding your comment and my naive reading of the patch:

Use -define webp:lossless=1 -quality 90, for example.

What is the relationship between webp:lossless and the webp encoder's near_lossless now? I see the ImageMagick WebP docs have removed the mention of webp:near-lossless since yesterday.

Furthermore, from reading coders/webp.c on main as of this patch I don't understand the interaction between webp:lossless, -quality, and the webp encoder's near_lossless.

[alanorth]

Thank you for the patch, @urban-warrior. From reading the coders/webp.c code, ImageMagick seems to always configure near_lossless mode using the given -quality if the cwebp encoder ABI supports it:

  if (image->quality != UndefinedCompressionQuality)
    {
      configure.quality=(float) image->quality;
#if WEBP_ENCODER_ABI_VERSION >= 0x020e
      configure.near_lossless=(float) image->quality;
#endif
    }

Thanks

[urban-warrior]

The theory is that the near lossless parameter is a no-op unless lossless compression is enabled. Our testing confirms our conjecture.