generate an anisotropic Gaussian kernel Args: ksize : e.g., 15, kernel size theta : [0, pi], rotation angle range l1

Args: x: HxWxC image, [0, 1] sf: down-scale factor Return: bicubicly downsampled LR image

complex multiplication Args: t1: NxCxHxWx2, complex tensor t2: NxCxHxWx2 Returns: output: NxCxHxWx2

myfun = @(block_struct) reshape(block_struct.data,m,1); x1 = blockproc(x1,[nr nc],myfun); x1 = reshape(x1,m,Nb); x1 = sum(x1,2);

complex multiplication t1: NxCxHxWx2 output: NxCxHxWx2

x: image, NxcxHxW k: kernel, cx1xhxw

x: image, NxcxHxW k: kernel, cx1xhxw

# psf: NxCxhxw # shape: [H,W] # otf: NxCxHxWx2

a: tensor NxCxWxHx2 sf: scale factor out: tensor NxCx(W/sf)x(H/sf)x2x(sf^2)

split a into sfxsf distinct blocks Args: a: NxCxWxH sf: split factor Returns: b: NxCx(W/sf)x(H/sf)x(sf^2)

split a into sfxsf distinct blocks Args: a: NxCxWxHx2 sf: split factor Returns: b: NxCx(W/sf)x(H/sf)x2x(sf^2)

python code from: https://github.com/ys-koshelev/nla_deblur/blob/90fe0ab98c26c791dcbdf231fe6f938fca80e2a0/boundaries.py Reducing bound

bgr version of rgb2ycbcr only_y: only return Y channel Input: uint8, [0, 255] float, [0, 1]

# complex's conjugation t: NxCxHxWx2 output: NxCxHxWx2

complex's conjugation Args: t: NxCxHxWx2 Returns: output: NxCxHxWx2

blur + downsampling Args: x: HxWxC image, [0, 1]/[0, 255] k: hxw, double sf: down-scale factor Return:

complex multiplication t1: NxCxHxWx2 output: NxCxHxWx2

bicubic downsampling + blur Args: x: HxWxC image, [0, 1] k: hxw, double sf: down-scale factor Return: d

python code from: https://github.com/ronaldosena/imagens-medicas-2/blob/40171a6c259edec7827a6693a93955de2bd39e76/Aulas/aula_2_-_uniform_fil

Smoothing filter

Disk filter

Args: x: 225x10000 matrix dim_pca: 15 Returns: pca_matrix: 15x225

Args: img: numpy image, WxH or WxHxC sf: scale factor Return: cropped image

Kai Zhang (github: https://github.com/cszn) 03/03/2019 # opt_fft_size.m # compute an optimal data length for Fourier transforms

Convert point-spread function to optical transfer function. otf = p2o(psf) computes the Fast Fourier Transform (FFT) of the point-spread

Convert point-spread function to optical transfer function. otf = p2o(psf) computes the Fast Fourier Transform (FFT) of the point-spread

Arguments :param input: tensor of shape :math:`(N, C_{\text{in}}, H, [W, D]))` :param padding: (tuple): m-elem tuple where m is the de

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

# modified version of https://github.com/assafshocher/BlindSR_dataset_generator # Kai Zhang # min_var = 0.175 * sf # variance of the

blur + bicubic downsampling Args: x: HxWxC image, [0, 1] k: hxw, double sf: down-scale factor Return: d

x: tensor image, NxCxWxH

s-fold upsampler Upsampling the spatial size by filling the new entries with zeros x: tensor image, NxCxWxH

s-fold upsampler Upsampling the spatial size by filling the new entries with zeros x: tensor image, NxCxWxH

Extends image to a certain size with zeros Parameters ---------- image: real 2d `numpy.ndarray` Input image shape:

Extends image to a certain size with zeros Parameters ---------- image: real 2d `numpy.ndarray` Input image shape:

x: image, NxcxHxW k: kernel, cx1xhxw sf: scale factor center: the first one or the moddle one Matlab function: tmp =

x: image, NxcxHxW k: kernel, cx1xhxw Matlab function: tmp = imfilter(x,h,'circular'); y = downsample2(tmp,K);

x: image, NxcxHxW k: kernel, cx1xhxw sf: scale factor center: the first one or the moddle one Matlab function: tmp =

x: image, NxcxHxW k: kernel, cx1xhxw Matlab function: tmp = upsample2(x,K); y = imfilter(tmp,h,'circular');

x1 = FB.*FR; FBR = BlockMM(nr,nc,Nb,m,x1); invW = BlockMM(nr,nc,Nb,m,F2B); invWBR = FBR./(invW + tau*Nb); fun = @(block_stru

FB: NxCxWxHx2 F2B: NxCxWxHx2 x1 = FB.*FR; FBR = BlockMM(nr,nc,Nb,m,x1); invW = BlockMM(nr,nc,Nb,m,F2B); invWBR = FBR.