Closed Ampflower closed 1 month ago
Describe the bug cjpegli's quality flag doesn't use the equivalent distance like cjxl does, instead using the distance of 1 regardless of input.
cjpegli
cjxl
To Reproduce
cjpegli input.png output.jpg -q 70
Expected behavior For the distance to be roughly 2.5, rather than 1.
Screenshots Images inapplicable as it is a logic problem. Instead, a log:
$ ./cjpegli input.png output.jpg -q 70 -v -v Read 1920x1080 image, 1856814 bytes. Encoding [YUV d1.000 AQ p2 OPT] Compressed to 203130 bytes (0.784 bpp). 1920 x 1080, 40.560 MP/s [40.56, 40.56], , 1 reps, 1 threads.
Environment
Additional context Happens regardless of any input. However, the -d flag works as expected, with changes in the log corresponding.
-d
$ ./cjpegli input.png output.jpg -d 3 -v -v Read 1920x1080 image, 1856814 bytes. Encoding [YUV d3.000 AQ p2 OPT] Compressed to 195866 bytes (0.756 bpp). 1920 x 1080, 40.999 MP/s [41.00, 41.00], , 1 reps, 1 threads.
I don't think the "d1.000" output there really means it's the same as -d 1. If you try different -q ## values, in practice the output size will be different, I presume.
"d1.000"
-d 1
-q ##
Describe the bug
cjpegli
's quality flag doesn't use the equivalent distance likecjxl
does, instead using the distance of 1 regardless of input.To Reproduce
cjpegli input.png output.jpg -q 70
Expected behavior For the distance to be roughly 2.5, rather than 1.
Screenshots Images inapplicable as it is a logic problem. Instead, a log:
Environment
Additional context Happens regardless of any input. However, the
-d
flag works as expected, with changes in the log corresponding.