QUOTE(-TT- @ Dec 4 2024, 18:40)
WebP looks pretty awful for color images honestly.
At least do an apples-to-apples comparison if you want to make statements like that.
Here's the same image converted to 1280x
PNG (lossless),
JPEG and
WebP using the legacy JPEG and current WebP quality values. Just open them in three tabs with the PNG in the middle and compare yourself. Not only is the WebP closer to the lossless version, it is almost
half the size of the JPEG (28kB vs 48kB).
As for the colors:
CODE
PNG -> WebP:
Channel distortion: RMSE
red: 648.953 (0.00990238)
green: 502.039 (0.00766063)
blue: 587.323 (0.00896197)
all: 582.561 (0.00888931)
Channel distortion: NCC
red: 0.998782
green: 0.998835
blue: 0.997796
all: 0.998471
Channel distortion: MAE
red: 431.974 (0.0065915)
green: 361.486 (0.00551592)
blue: 407.974 (0.00622528)
all: 400.478 (0.0061109)
PNG -> JPEG:
Channel distortion: RMSE
red: 671.641 (0.0102486)
green: 539.765 (0.00823628)
blue: 647.792 (0.00988467)
all: 622.383 (0.00949695)
Channel distortion: NCC
red: 0.99871
green: 0.998552
blue: 0.997448
all: 0.998237
Channel distortion: MAE
red: 440.627 (0.00672354)
green: 368.067 (0.00561634)
blue: 452.54 (0.00690531)
all: 420.411 (0.00641507)
The WebP resample has less deviance on literally every color channel with every metric. (For Root Mean Squared Error and Mean Absolute Error, smaller is better. For Normalized Cross Correlation, larger is better.)
And while there are some parts that look sharper on the JPEG, there is also much more visible artifacting. You are free to prefer the JPEG if you wish, but you're not going to convince me that it's worth twice the filesize or even that it provides a better overall result.