Some metrics comparison:

2048x1320_nitish-kadam-34748

Butteraugli(AVIF - HEIC - HTJ2K - WebP - JPEG XL - MozJpeg)

3-norm:(AVIF - HEIC - HTJ2K - JPEG XL - WebP - MozJpeg)

AVIF9.2743740082 3-norm: 2.321073

HEIC13.8783264160 3-norm: 3.060689

HTJ2K17.1095638275 3-norm: 4.099578

MozJpeg19.1906070709 3-norm: 8.023822

JPEG XL19.0907402039 3-norm: 4.272680

WebP17.9173774719 3-norm: 5.007857(AVIF - HEIC - JPEG XL - HTJ2K - WebP - MozJpeg)

DSSIM

0.007985AVIF

0.011601HEIC

0.017184HTJ2K

0.086505MozJpeg

0.015210JPEG XL

0.031527WebPSSIMULACRA(AVIF - HEIC - JPEG XL - HTJ2K - WebP - MozJpeg)

0.02931092493236065AVIF

0.038233496248722076HEIC

0.05292587727308273HTJ2K

0.10630584508180618MozJpeg

0.05085379257798195JPEG XL

0.06478796154260635WebPVMAF:(AVIF - HEIC - HTJ2K - JPEG XL - WebP - MozJpeg)

AVIFVMAF: 87.43534236386863, PSNR: 44.77028232276835, SSIM: 0.9923002123832703, MS-SSIM: 0.9906416878743063

HEICVMAF: 80.8764654659456, PSNR: 42.94141442285002, SSIM: 0.9905341863632202, MS-SSIM: 0.9884282902889149

HTJ2KVMAF: 73.2913217099848, PSNR: 40.91610359405568, SSIM: 0.9820453524589539, MS-SSIM: 0.9816313764353033

MozJpegVMAF: 54.10929529591918, PSNR: 35.5303285917693, SSIM: 0.9309688806533813, MS-SSIM: 0.936962545830246

JPEG XLVMAF: 61.59250694065548, PSNR: 39.93377021739749, SSIM: 0.9857387542724609, MS-SSIM: 0.9829265315009978

WebPVMAF: 55.94460452312931, PSNR: 37.98409101496202, SSIM: 0.969596803188324, MS-SSIM: 0.9705516740404905

But, on a larger bpp (in other examples, the result is almost similar):

2048x1320_alex-siale-95113(~506000 bytes each image)

Butteraugli(JPEG XL - MozJpeg - HEIC - AVIF - HTJ2K - WebP)

3-norm:(JPEG XL - MozJpeg - HEIC - WebP - HTJ2K - AVIF)

AVIF8.8833646774 3-norm: 4.289337

HEIC8.6990165710 3-norm: 3.963638

HTJ2K9.4587421417 3-norm: 4.243970

MozJpeg7.4712114334 3-norm: 3.394969

JPEG XL1.8749873638 3-norm: 1.050182

WebP9.5613374710 3-norm: 3.975210DSSIM(JPEG XL - HEIC - AVIF - MozJpeg - WebP - HTJ2K)

0.011497AVIF

0.011003HEIC

0.017835HTJ2K

0.012662MozJpeg

0.009563JPEG XL

0.014941WebPSSIMULACRA(JPEG XL - HEIC - AVIF - WebP - HTJ2K - MozJpeg)

0.03911368176341057AVIF

0.03868887946009636HEIC

0.05217859894037247HTJ2K

0.05393524467945099MozJpeg

0.035159382969141006JPEG XL

0.04311709105968475WebPVMAF: (MozJpeg - HTJ2K - HEIC - AVIF - WebP - JPEG XL)

SSIM: (MozJpeg - HEIC - AVIF - WebP - HTJ2K - JPEG XL)

AVIFVMAF: 91.62228049091999, PSNR: 37.9803197776212, SSIM: 0.9977059364318848, MS-SSIM: 0.9922499111306672

HEICVMAF: 92.00663265691058, PSNR: 38.64841753304782, SSIM: 0.9986478090286255, MS-SSIM: 0.9948372815741532

HTJ2KVMAF: 92.19011608093165, PSNR: 36.41784272626444, SSIM: 0.9966325163841248, MS-SSIM: 0.9897441535051162

MozJpegVMAF: 92.73296699245252, PSNR: 34.54616417406431, SSIM: 0.9990101456642151, MS-SSIM: 0.9928436494191371

JPEG XLVMAF: 89.07241730929319, PSNR: 36.8834302952345, SSIM: 0.9963483810424805, MS-SSIM: 0.9906952076528491

WebPVMAF: 90.0164064491984, PSNR: 36.00095502257268, SSIM: 0.9976305365562439, MS-SSIM: 0.9916054652684111

2048x1320_andrew-coelho-46449(~432000 bytes each image)

Butteraugli,DSSIM,SSIMULACRA(JPEG XL - best ... HTJ2K - worst)

VMAF,SSIM:(MozJpeg - best ... JPEG XL - worst) - For example, Netflix uses these metrics in the Framework and on the blog (with a note that VMAF is more suitable for video).