calculate SSIM the same outputs as MATLAB's img1, img2: [0, 255]

# complex's conjugation t: NxCxHxWx2 output: NxCxHxWx2

# x[N, 1, W, H] -> x[N, 1, W + 2 pad, H + 2 pad] (pariodic padding)

s-fold downsampler Keeping the upper-left pixel for each distinct sfxsf patch and discarding the others x: tensor image, NxCxWxH

s-fold downsampler Keeping the upper-left pixel for each distinct sfxsf patch and discarding the others x: tensor image, NxCxWxH

x: tensor, NxCxWxH k: tensor, Nx(1,3)xwxh sf: integer, 1 sigma: tensor, Nx1x1x1

x: tensor, NxCxWxH k: tensor, Nx(1,3)xwxh sf: integer, 1 sigma: tensor, Nx1x1x1

:param x: (b, c, t, h, w) :return:

img: HxWxC kernel: hxw denominator: HxWx1 upperleft: HxWxC

img: NxCxHxW kernel: Nx1xhxw denominator: Nx1xHxW upperleft: NxCxHxWx2

set up logger

img: tensor image, NxCxWxH or CxWxH or WxH sf: scale factor

Args: psf: NxCxhxw shape: [H,W] Returns: otf: NxCxHxWx2

Convert point-spread function to optical transfer function. Compute the Fast Fourier Transform (FFT) of the point-spread function (PSF

same as matlab rgb2ycbcr only_y: only return Y channel Input: uint8, [0, 255] float, [0, 1]

shift pixel for super-resolution with different scale factors Args: x: WxHxC or WxH, image or kernel sf: scale factor

x: PCA projection matrix, e.g., 15x225

Converts a torch Tensor into an image Numpy array of BGR channel order Input: 4D(B,(3/1),H,W), 3D(C,H,W), or 2D(H,W), any range, RGB channel