- From: John Bowler <john.cunningham.bowler@gmail.com>
- Date: Fri, 9 May 2025 09:40:19 -0700
- To: "Chris Blume (ProgramMax)" <programmax@gmail.com>
- Cc: public-png@w3.org
>> So in that limiting test using an image which is quite noisy the >> filter restriction cost me about 1% in compression. >That is WAYY better than I expected. >In my head, I was imagining people weighing 10% file size for 4x decode speed. Granted, it is a large image. But not *that* large today. And we aren't resetting the compressor yet. This is great news! Unfortunately not; my test was garbage because oxipng apparently ignores the filters given in the --filters argument. pngcheck -vv shows that all my output files use all the filters. Darn. The original image and all the oxipng outputs use AVG almost exclusively. Given the noise in the image this makes sense; the average of the three adjacent pixel values will have must less noise than any single one of them. AVG is the only arithmetic filter;png all the others use a single adjacent pixel value. More work required :-) John Bowler
Received on Friday, 9 May 2025 16:40:34 UTC