pyiqa.archs.vif_arch

VIF Metric

@article{sheikh2006vif,

title={Image information and visual quality}, author={Sheikh, Hamid R and Bovik, Alan C}, journal={IEEE Transactions on image processing}, volume={15}, number={2}, pages={430–444}, year={2006}, publisher={IEEE}

}

Created by: https://github.com/dingkeyan93/IQA-optimization/blob/master/IQA_pytorch/VIF.py

Modified by: Jiadi Mo (https://github.com/JiadiMo)

Refer to:

Matlab code from http://live.ece.utexas.edu/research/Quality/vifvec_release.zip;

Module Contents

pyiqa.archs.vif_arch.sp5_filters()

Define spatial filters.

pyiqa.archs.vif_arch.corrDn(image, filt, step=1, channels=1)

Compute correlation of image with FILT, followed by downsampling. :param image: A tensor. Shape \((N, C, H, W)\). :param filt: A filter. :param step: Downsampling factors. :param channels: Number of channels.

pyiqa.archs.vif_arch.SteerablePyramidSpace(image, height=4, order=5, channels=1)

Construct a steerable pyramid on image. :param image: A tensor. Shape \((N, C, H, W)\). :param height: Number of pyramid levels to build. :type height: int :param order: Number of orientations. :type order: int :param channels: Number of channels. :type channels: int

class pyiqa.archs.vif_arch.VIF(channels=1, level=4, ori=6)

Bases: torch.nn.Module

Image Information and Visual Quality metric :param channels: Number of channels. :type channels: int :param level: Number of levels to build. :type level: int :param ori: Number of orientations. :type ori: int

Reference:

Sheikh, Hamid R., and Alan C. Bovik. “Image information and visual quality.” IEEE Transactions on image processing 15, no. 2 (2006): 430-444.

corrDn_win(image, filt, step=1, channels=1, start=[0, 0], end=[0, 0])

Compute correlation of image with FILT using window, followed by downsampling. :param image: A tensor. Shape \((N, C, H, W)\). :param filt: A filter. :param step: Downsampling factors. :type step: int :param channels: Number of channels. :type channels: int :param start: The window over which the convolution occurs. :type start: list :param end: The window over which the convolution occurs. :type end: list

vifsub_est_M(org, dist)

Calculate the parameters of the distortion channel. :param org: A reference tensor. Shape \((N, C, H, W)\). :param dist: A distortion tensor. Shape \((N, C, H, W)\).

refparams_vecgsm(org)

Calculate the parameters of the reference image. :param org: A reference tensor. Shape \((N, C, H, W)\).

vif(x, y)

VIF metric. Order of input is important. :param x: A distortion tensor. Shape \((N, C, H, W)\). :param y: A reference tensor. Shape \((N, C, H, W)\).

forward(X, Y)

Args: x: A distortion tensor. Shape \((N, C, H, W)\). y: A reference tensor. Shape \((N, C, H, W)\). Order of input is important.