Is your feature request related to a problem? Please describe.
Proposed in 2017 by Golestaneh et al in Synthesized Texture Quality Assessment via Multi-scale Spatial and Statistical Texture Attributes of Image and Gradient Magnitude Coefficients, it's a good metric to quantify the structural perceptual similarity between two images. E.g., an original image, and a synthesized procedural image.
Describe the solution you'd like
An implementation in pytorch would be excellent, and it could complement LPIPS and DISTS metrics as far as perceptual metrics go.
Describe alternatives you've considered
I've considered LPIPS, DISTS, and was considering STSIM, structural texture similarity image metric. Together they might give a more accurate portrait of structural perceptual similarity between natural textures, and synthesized procedural textures for example.
Additional context
The original paper can be found here (PDF file, Arxiv)
and a MatLab implementation by the authors can be found here on github
Is your feature request related to a problem? Please describe. Proposed in 2017 by Golestaneh et al in Synthesized Texture Quality Assessment via Multi-scale Spatial and Statistical Texture Attributes of Image and Gradient Magnitude Coefficients, it's a good metric to quantify the structural perceptual similarity between two images. E.g., an original image, and a synthesized procedural image.
Describe the solution you'd like An implementation in pytorch would be excellent, and it could complement LPIPS and DISTS metrics as far as perceptual metrics go.
Describe alternatives you've considered I've considered LPIPS, DISTS, and was considering STSIM, structural texture similarity image metric. Together they might give a more accurate portrait of structural perceptual similarity between natural textures, and synthesized procedural textures for example.
Additional context The original paper can be found here (PDF file, Arxiv) and a MatLab implementation by the authors can be found here on github