Waifu2x/utils/image_quality.py

# Pytorch Multi-Scale Structural Similarity Index (SSIM)
# This code is written by jorge-pessoa (https://github.com/jorge-pessoa/pytorch-msssim)
# MIT licence
import math
from math import exp

import torch
import torch.nn.functional as F
from torch.autograd import Variable


# +++++++++++++++++++++++++++++++++++++
#           SSIM
# -------------------------------------


def gaussian(window_size, sigma):
    gauss = torch.Tensor(
        [
            exp(-((x - window_size // 2) ** 2) / float(2 * sigma**2))
            for x in range(window_size)
        ]
    )
    return gauss / gauss.sum()


def create_window(window_size, channel):
    _1D_window = gaussian(window_size, 1.5).unsqueeze(1)
    _2D_window = _1D_window.mm(_1D_window.t()).float().unsqueeze(0).unsqueeze(0)
    window = Variable(
        _2D_window.expand(channel, 1, window_size, window_size).contiguous()
    )
    return window


def _ssim(img1, img2, window, window_size, channel, size_average=True, full=False):
    padd = 0

    mu1 = F.conv2d(img1, window, padding=padd, groups=channel)
    mu2 = F.conv2d(img2, window, padding=padd, groups=channel)

    mu1_sq = mu1.pow(2)
    mu2_sq = mu2.pow(2)
    mu1_mu2 = mu1 * mu2

    sigma1_sq = F.conv2d(img1 * img1, window, padding=padd, groups=channel) - mu1_sq
    sigma2_sq = F.conv2d(img2 * img2, window, padding=padd, groups=channel) - mu2_sq
    sigma12 = F.conv2d(img1 * img2, window, padding=padd, groups=channel) - mu1_mu2

    C1 = 0.01**2
    C2 = 0.03**2

    ssim_map = ((2 * mu1_mu2 + C1) * (2 * sigma12 + C2)) / (
        (mu1_sq + mu2_sq + C1) * (sigma1_sq + sigma2_sq + C2)
    )

    v1 = 2.0 * sigma12 + C2
    v2 = sigma1_sq + sigma2_sq + C2
    cs = torch.mean(v1 / v2)

    if size_average:
        ret = ssim_map.mean()
    else:
        ret = ssim_map.mean(1).mean(1).mean(1)

    if full:
        return ret, cs
    return ret


class SSIM(torch.nn.Module):
    def __init__(self, window_size=11, size_average=True):
        super(SSIM, self).__init__()
        self.window_size = window_size
        self.size_average = size_average
        self.channel = 1
        self.window = create_window(window_size, self.channel)

    def forward(self, img1, img2):
        (_, channel, _, _) = img1.size()

        if channel == self.channel and self.window.data.type() == img1.data.type():
            window = self.window
        else:
            window = create_window(self.window_size, channel)

            if img1.is_cuda:
                window = window.cuda(img1.get_device())
            window = window.type_as(img1)

            self.window = window
            self.channel = channel

        return _ssim(img1, img2, window, self.window_size, channel, self.size_average)


def ssim(img1, img2, window_size=11, size_average=True, full=False):
    (_, channel, height, width) = img1.size()

    real_size = min(window_size, height, width)
    window = create_window(real_size, channel)

    if img1.is_cuda:
        window = window.cuda(img1.get_device())
    window = window.type_as(img1)

    return _ssim(img1, img2, window, real_size, channel, size_average, full=full)


def msssim(img1, img2, window_size=11, size_average=True):
    # TODO: fix NAN results
    if img1.size() != img2.size():
        raise RuntimeError(
            "Input images must have the same shape (%s vs. %s)."
            % (img1.size(), img2.size())
        )
    if len(img1.size()) != 4:
        raise RuntimeError(
            "Input images must have four dimensions, not %d" % len(img1.size())
        )

    weights = torch.tensor([0.0448, 0.2856, 0.3001, 0.2363, 0.1333], dtype=img1.dtype)
    if img1.is_cuda:
        weights = weights.cuda(img1.get_device())

    levels = weights.size()[0]
    mssim = []
    mcs = []
    for _ in range(levels):
        sim, cs = ssim(
            img1, img2, window_size=window_size, size_average=size_average, full=True
        )
        mssim.append(sim)
        mcs.append(cs)

        img1 = F.avg_pool2d(img1, (2, 2))
        img2 = F.avg_pool2d(img2, (2, 2))

    mssim = torch.stack(mssim)
    mcs = torch.stack(mcs)
    return torch.prod(mcs[0 : levels - 1] ** weights[0 : levels - 1]) * (
        mssim[levels - 1] ** weights[levels - 1]
    )


class MSSSIM(torch.nn.Module):
    def __init__(self, window_size=11, size_average=True, channel=3):
        super(MSSSIM, self).__init__()
        self.window_size = window_size
        self.size_average = size_average
        self.channel = channel

    def forward(self, img1, img2):
        # TODO: store window between calls if possible
        return msssim(
            img1, img2, window_size=self.window_size, size_average=self.size_average
        )


def calc_psnr(sr, hr, scale=0, benchmark=False):
    # adapt from EDSR: https://github.com/thstkdgus35/EDSR-PyTorch
    diff = (sr - hr).data
    if benchmark:
        shave = scale
        if diff.size(1) > 1:
            convert = diff.new(1, 3, 1, 1)
            convert[0, 0, 0, 0] = 65.738
            convert[0, 1, 0, 0] = 129.057
            convert[0, 2, 0, 0] = 25.064
            diff.mul_(convert).div_(256)
            diff = diff.sum(dim=1, keepdim=True)
    else:
        shave = scale + 6

    valid = diff[:, :, shave:-shave, shave:-shave]
    mse = valid.pow(2).mean()

    return -10 * math.log10(mse)


# +++++++++++++++++++++++++++++++++++++
#           PSNR
# -------------------------------------
from torch import nn


def psnr(predict, target):
    with torch.no_grad():
        criteria = nn.MSELoss()
        mse = criteria(predict, target)
        return -10 * torch.log10(mse)