The following tools were tested on PNG test corpus by Cédric Louvrier and Guetzli photo dataset (converted to JPEG at highest quality) by Jyrki Alakuijala.
|PNG||273 711 B||296 230 B||–||–||–||304 357 B||–|
|JPEG||16 861 910 B||16 862 728 B||–||–||–||–||–|
|Visually lossless optimization|
|PNG||268 084 B||297 979 B||–||–||217 389 B||–||–|
|JPEG||3 717 075 B||2 083 557 B||4 207 455 B||3 738 251 B||3 579 727 B||3 629 611 B||2 676 969 B|
*Score shows how many images are visually indistinguishable from originals under normal viewing conditions. Additional tests revealed other issues.
- ImageOptim scored 24/24 in lossy PNG compression but did not perform well on gradients and flat graphics in extended testing.
- Tools like TinyPNG and Crunch use the original PNGQuant for color quantization, and generally perform worse than ImageOptim in terms of quality and lossless compression.
- TinyPNG images could be losslessly compressed by 6.4% more using Optimage.
- Guetzli achieved best quality/size ratio in some cases but failed to provide consistent quality especially at lower quality settings.
- Guetzli is extremely slow requiring approximately 1 minute of CPU and 100 MB of memory per 1 MPix.