a  dn@sddlZddlZddlZddlZddlZddlmZddlm Z ddl m Z ddZ ddZ d d Zd d ZdNddZdOddZdPddZdQddZdRddZdSddZdddej ejfdddfddZejddd dTd!d"ZdUd%d&ZdVd'd(ZdWd)d*ZdXd+d,ZdYd.d/ZdZd1d2Zd[d3d4Z d\d5d6Z!d]d8d9Z"d^d:d;Z#d_dd?Z%dad@dAZ&dbdBdCZ'dcdDdEZ(dddFdGZ)dedIdJZ*dfdLdMZ+dS)gN)special)multivariate_normal)rgb_to_grayscalecCsdt|ddgd|dgg}tt|t| gt|t|gg}t|t||jS)zCalculate the rotated sigma matrix (two dimensional matrix). Args: sig_x (float): sig_y (float): theta (float): Radian measurement. Returns: ndarray: Rotated sigma matrix. r)nparraycossindotT)sig_xsig_ythetad_matrixZu_matrixrb/var/www/html/stable-diffusion-webui/venv/lib/python3.9/site-packages/basicsr/data/degradations.py sigma_matrix2s 0rcCsjt| dd|dd}t||\}}t|||df|||df||d}|||fS)a&Generate the mesh grid, centering at zero. Args: kernel_size (int): Returns: xy (ndarray): with the shape (kernel_size, kernel_size, 2) xx (ndarray): with the shape (kernel_size, kernel_size) yy (ndarray): with the shape (kernel_size, kernel_size) r?)rZarangeZmeshgridZhstackZreshape) kernel_sizeaxxxyyZxyrrr mesh_grid s rc Cs2tj|}tdtt|||d}|S)a.Calculate PDF of the bivariate Gaussian distribution. Args: sigma_matrix (ndarray): with the shape (2, 2) grid (ndarray): generated by :func:`mesh_grid`, with the shape (K, K, 2), K is the kernel size. Returns: kernel (ndarrray): un-normalized kernel. r)rlinalginvexpsumr ) sigma_matrixgrid inverse_sigmakernelrrrpdf22s "r#cCs4tddgddgddgg}t||}||}|S)aOCalculate the CDF of the standard bivariate Gaussian distribution. Used in skewed Gaussian distribution. Args: d_matrix (ndarrasy): skew matrix. grid (ndarray): generated by :func:`mesh_grid`, with the shape (K, K, 2), K is the kernel size. Returns: cdf (ndarray): skewed cdf. rr)rrr cdf)rr rvr$rrrcdf2Bs   r&Tc Csb|durt|\}}}|r:t|ddgd|dgg}n t|||}t||}|t|}|S)aGenerate a bivariate isotropic or anisotropic Gaussian kernel. In the isotropic mode, only `sig_x` is used. `sig_y` and `theta` is ignored. Args: kernel_size (int): sig_x (float): sig_y (float): theta (float): Radian measurement. grid (ndarray, optional): generated by :func:`mesh_grid`, with the shape (K, K, 2), K is the kernel size. Default: None isotropic (bool): Returns: kernel (ndarray): normalized kernel. Nrr)rrrrr#r) rr r rr isotropic_rr"rrrbivariate_GaussianTs   r)c Cs|durt|\}}}|r:t|ddgd|dgg}n t|||}tj|} tdttt || |d|} | t| } | S)aGenerate a bivariate generalized Gaussian kernel. Described in `Parameter Estimation For Multivariate Generalized Gaussian Distributions`_ by Pascal et. al (2013). In the isotropic mode, only `sig_x` is used. `sig_y` and `theta` is ignored. Args: kernel_size (int): sig_x (float): sig_y (float): theta (float): Radian measurement. beta (float): shape parameter, beta = 1 is the normal distribution. grid (ndarray, optional): generated by :func:`mesh_grid`, with the shape (K, K, 2), K is the kernel size. Default: None Returns: kernel (ndarray): normalized kernel. .. _Parameter Estimation For Multivariate Generalized Gaussian Distributions: https://arxiv.org/abs/1302.6498 Nrrr) rrrrrrrpowerrr rr r rbetar r'r(rr!r"rrrbivariate_generalized_Gaussianps   *r-c Cs|durt|\}}}|r:t|ddgd|dgg}n t|||}tj|} tttt || |d|d} | t| } | S)aGenerate a plateau-like anisotropic kernel. 1 / (1+x^(beta)) Ref: https://stats.stackexchange.com/questions/203629/is-there-a-plateau-shaped-distribution In the isotropic mode, only `sig_x` is used. `sig_y` and `theta` is ignored. Args: kernel_size (int): sig_x (float): sig_y (float): theta (float): Radian measurement. beta (float): shape parameter, beta = 1 is the normal distribution. grid (ndarray, optional): generated by :func:`mesh_grid`, with the shape (K, K, 2), K is the kernel size. Default: None Returns: kernel (ndarray): normalized kernel. Nrrr) rrrrrrZ reciprocalr*rr r+rrrbivariate_plateaus   *r.c Cs|ddksJd|d|dks,Jdtj|d|d}|dur|d|dksbJd|d|dkszJdtj|d|d}tj|d|d}n|}d}t|||||d } |d ur|d|dksJd tjj|d|d| jd } | | } | t| } | S) aRandomly generate bivariate isotropic or anisotropic Gaussian kernels. In the isotropic mode, only `sigma_x_range` is used. `sigma_y_range` and `rotation_range` is ignored. Args: kernel_size (int): sigma_x_range (tuple): [0.6, 5] sigma_y_range (tuple): [0.6, 5] rotation range (tuple): [-math.pi, math.pi] noise_range(tuple, optional): multiplicative kernel noise, [0.75, 1.25]. Default: None Returns: kernel (ndarray): rr"Kernel size must be an odd number.rWrong sigma_x_range.FWrong sigma_y_range.Wrong rotation_range.r'NWrong noise range.size)rrandomuniformr)shaper) r sigma_x_range sigma_y_rangerotation_range noise_ranger'sigma_xsigma_yrotationr"noiserrrrandom_bivariate_Gaussians" rBc CsR|ddksJd|d|dks,Jdtj|d|d}|dur|d|dksbJd|d|dkszJdtj|d|d}tj|d|d} n|}d} tjd krtj|dd} ntjd|d} t|||| | |d } |d ur@|d|dksJd tjj|d|d| jd } | | } | t| } | S)a Randomly generate bivariate generalized Gaussian kernels. In the isotropic mode, only `sigma_x_range` is used. `sigma_y_range` and `rotation_range` is ignored. Args: kernel_size (int): sigma_x_range (tuple): [0.6, 5] sigma_y_range (tuple): [0.6, 5] rotation range (tuple): [-math.pi, math.pi] beta_range (tuple): [0.5, 8] noise_range(tuple, optional): multiplicative kernel noise, [0.75, 1.25]. Default: None Returns: kernel (ndarray): rrr/rr0Fr1r2?r3Nr4r5)rr7r8r-r9r rr:r;r<Z beta_ranger=r'r>r?r@r,r"rArrr%random_bivariate_generalized_Gaussians( rEc CsR|ddksJd|d|dks,Jdtj|d|d}|dur|d|dksbJd|d|dkszJdtj|d|d}tj|d|d} n|}d} tjd krtj|dd} ntjd|d} t|||| | |d } |d ur@|d|dksJd tjj|d|d| jd } | | } | t| } | S)aRandomly generate bivariate plateau kernels. In the isotropic mode, only `sigma_x_range` is used. `sigma_y_range` and `rotation_range` is ignored. Args: kernel_size (int): sigma_x_range (tuple): [0.6, 5] sigma_y_range (tuple): [0.6, 5] rotation range (tuple): [-math.pi/2, math.pi/2] beta_range (tuple): [1, 4] noise_range(tuple, optional): multiplicative kernel noise, [0.75, 1.25]. Default: None Returns: kernel (ndarray): rrr/rr0Fr1r2rCr3Nr4r5)rr7r8r.r9rrDrrrrandom_bivariate_plateaus( rF)g333333?)rCc Cst||d} | dkr.t|||||dd} n| dkrLt|||||dd} n~| dkrlt||||||dd} n^| dkrt||||||dd} n>| d krt|||||d dd} n| d krt|||||d dd} | S) aRandomly generate mixed kernels. Args: kernel_list (tuple): a list name of kernel types, support ['iso', 'aniso', 'skew', 'generalized', 'plateau_iso', 'plateau_aniso'] kernel_prob (tuple): corresponding kernel probability for each kernel type kernel_size (int): sigma_x_range (tuple): [0.6, 5] sigma_y_range (tuple): [0.6, 5] rotation range (tuple): [-math.pi, math.pi] beta_range (tuple): [0.5, 8] noise_range(tuple, optional): multiplicative kernel noise, [0.75, 1.25]. Default: None Returns: kernel (ndarray): rZisoT)r=r'ZanisoFZgeneralized_isoZgeneralized_anisoZ plateau_isoNZ plateau_aniso)r7choicesrBrErF) Z kernel_listZ kernel_probrr:r;r<Z betag_rangeZ betap_ranger=Z kernel_typer"rrrrandom_mixed_kernelsGsL  rKignore)divideinvalidcsddksJdtfddg}ddtj|ddddf<|t|}|kr|d}t|||f||ff}|S)aM2D sinc filter, ref: https://dsp.stackexchange.com/questions/58301/2-d-circularly-symmetric-low-pass-filter Args: cutoff (float): cutoff frequency in radians (pi is max) kernel_size (int): horizontal and vertical size, must be odd. pad_to (int): pad kernel size to desired size, must be odd or zero. rrr/c sttt|ddd|ddddtjt|ddd|dddS)Nrr)rZj1rsqrtpi)xycutoffrrrs & &z)circular_lowpass_kernel..)rZ fromfunctionrPrpad)rTrZpad_tor"Zpad_sizerrSrcircular_lowpass_kernels * rX FcCsb|rBttjj|jdd|d}tj|ddjddd}nttjj|j|d}|S)a/Generate Gaussian noise. Args: img (Numpy array): Input image, shape (h, w, c), range [0, 1], float32. sigma (float): Noise scale (measured in range 255). Default: 10. Returns: (Numpy array): Returned noisy image, shape (h, w, c), range[0, 1], float32. rro@Zaxis)rfloat32r7randnr9Z expand_dimsrepeat)imgsigma gray_noiserArrrgenerate_gaussian_noises $rccCsdt|||}||}|r8|r8t|dddd}n(|rLt|dd}n|r`|dd}|S)a*Add Gaussian noise. Args: img (Numpy array): Input image, shape (h, w, c), range [0, 1], float32. sigma (float): Noise scale (measured in range 255). Default: 10. Returns: (Numpy array): Returned noisy image, shape (h, w, c), range[0, 1], float32. rZrr)rcrclipround)r`rareroundsrbrAoutrrradd_gaussian_noises ric Cs|\}}}}t|ttfs4||dddd}t|ttfrL|dk}n||ddd}t|dk}|rtj|dd|j|j d|d}||d||}tj||j|j d|d} |r| d|||} | S)Add Gaussian noise (PyTorch version). Args: img (Tensor): Shape (b, c, h, w), range[0, 1], float32. scale (float | Tensor): Noise scale. Default: 1.0. Returns: (Tensor): Returned noisy image, shape (b, c, h, w), range[0, 1], float32. rrrrVdtypedevicerZ) r6 isinstancefloatintviewtorchrr^rlrm) r`rarbbr(hwcal_gray_noise noise_grayrArrrgenerate_gaussian_noise_pts  *"rxcCsdt|||}||}|r8|r8t|dddd}n(|rLt|dd}n|r`|dd}|S)rjrZrrdr)rxrrclamprf)r`rarbrergrArhrrradd_gaussian_noise_pts rzrrYcCs:tj|d|d}tj|kr*d}nd}t|||SNrrTF)rr7r8rcr` sigma_range gray_probrarbrrrrandom_generate_gaussian_noises rrrcCsdt|||}||}|r8|r8t|dddd}n(|rLt|dd}n|r`|dd}|SNrZrrdr)rrrerfr`r~rrergrArhrrrrandom_add_gaussian_noise s rcCsdtj|d|j|jd|d|d|d}tj|d|j|jd}||k}t|||SNrrkr)rrrandr6rlrmrorxr}rrr!random_generate_gaussian_noise_pts rcCsdt|||}||}|r8|r8t|dddd}n(|rLt|dd}n|r`|dd}|Sr)rrrryrfrrrrrandom_add_gaussian_noise_pt!s rrcCs|rt|tj}t|dddd}tt|}dtt |}t tj ||t |}||}|rtj|ddddtjfddd}||S)aGenerate poisson noise. Ref: https://github.com/scikit-image/scikit-image/blob/main/skimage/util/noise.py#L37-L219 Args: img (Numpy array): Input image, shape (h, w, c), range [0, 1], float32. scale (float): Noise scale. Default: 1.0. gray_noise (bool): Whether generate gray noise. Default: False. Returns: (Numpy array): Returned noisy image, shape (h, w, c), range[0, 1], float32. rZrrdrNr\r[)cv2ZcvtColorZCOLOR_BGR2GRAYrrerflenuniqueceillog2r]r7poissonror_Znewaxis)r`scalerbvalsrhrArrrgenerate_poisson_noise0s$rcCsdt|||}||}|r8|r8t|dddd}n(|rLt|dd}n|r`|dd}|S)aZAdd poisson noise. Args: img (Numpy array): Input image, shape (h, w, c), range [0, 1], float32. scale (float): Noise scale. Default: 1.0. gray_noise (bool): Whether generate gray noise. Default: False. Returns: (Numpy array): Returned noisy image, shape (h, w, c), range[0, 1], float32. rZrrdr)rrrerfr`rrergrbrArhrrradd_poisson_noiseKs rc s\}}}}t|ttfr(|dk}n||ddd}t|dk}|rtddtd dddfddt |D}dd|D} ||ddd} t | | } | } | |d ||} td dddfd dt |D}d d|D} ||ddd} t | | } | } |r\| d|| |} t|ttfs|||ddd}| |S) aGenerate a batch of poisson noise (PyTorch version) Args: img (Tensor): Input image, shape (b, c, h, w), range [0, 1], float32. scale (float | Tensor): Noise scale. Number or Tensor with shape (b). Default: 1.0. gray_noise (float | Tensor): 0-1 number or Tensor with shape (b). 0 for False, 1 for True. Default: 0. Returns: (Tensor): Returned noisy image, shape (b, c, h, w), range[0, 1], float32. rr)Znum_output_channelsrZrdc s2g|]*}tt|ddddddfqSNrrrr.0i)img_grayrr {z-generate_poisson_noise_pt..cSs g|]}dtt|qSrrrrrrrrrr|rr\c s2g|]*}tt|ddddddfqSrrr)r`rrrrcSs g|]}dtt|qSrrrrrrrr)r6rnrorprqrrrrryrfrangeZ new_tensorrexpand) r`rrbrsr(rtrurvZ vals_listrrhrwrAr)r`rrgenerate_poisson_noise_ptbs2  rcCsdt|||}||}|r8|r8t|dddd}n(|rLt|dd}n|r`|dd}|S)aAdd poisson noise to a batch of images (PyTorch version). Args: img (Tensor): Input image, shape (b, c, h, w), range [0, 1], float32. scale (float | Tensor): Noise scale. Number or Tensor with shape (b). Default: 1.0. gray_noise (float | Tensor): 0-1 number or Tensor with shape (b). 0 for False, 1 for True. Default: 0. Returns: (Tensor): Returned noisy image, shape (b, c, h, w), range[0, 1], float32. rZrrdr)rrrryrfrrrradd_poisson_noise_pts rcCs:tj|d|d}tj|kr*d}nd}t|||Sr|)rr7r8rr` scale_rangerrrbrrrrandom_generate_poisson_noises rcCsdt|||}||}|r8|r8t|dddd}n(|rLt|dd}n|r`|dd}|Sr)rrrerfr`rrrergrArhrrrrandom_add_poisson_noises rcCsdtj|d|j|jd|d|d|d}tj|d|j|jd}||k}t|||Sr)rrrr6rlrmrorrrrr random_generate_poisson_noise_pts rcCsdt|||}||}|r8|r8t|dddd}n(|rLt|dd}n|r`|dd}|Sr)rrrryrfrrrrrandom_add_poisson_noise_pts rZcCsLt|dd}ttj|g}td|d|\}}tt|dd}|S)agAdd JPG compression artifacts. Args: img (Numpy array): Input image, shape (h, w, c), range [0, 1], float32. quality (float): JPG compression quality. 0 for lowest quality, 100 for best quality. Default: 90. Returns: (Numpy array): Returned image after JPG, shape (h, w, c), range[0, 1], float32. rrz.jpgrZ)rrerprZIMWRITE_JPEG_QUALITYZimencoder]Zimdecode)r`qualityZ encode_paramr(Zencimgrrradd_jpg_compressions rrdcCs tj|d|d}t||S)aRandomly add JPG compression artifacts. Args: img (Numpy array): Input image, shape (h, w, c), range [0, 1], float32. quality_range (tuple[float] | list[float]): JPG compression quality range. 0 for lowest quality, 100 for best quality. Default: (90, 100). Returns: (Numpy array): Returned image after JPG, shape (h, w, c), range[0, 1], float32. rr)rr7r8r)r`Z quality_rangerrrrrandom_add_jpg_compressions r)NT)NT)NT)NT)NT)NT)r)rYF)rYTFF)rYr)rYrTF)r{r)rrTF)r{r)rrTF)rF)rTFF)rr)rTFr)rr)rrTF)rr)rrTF)r)r),rmathnumpyrr7rrZscipyrZ scipy.statsrZ(torchvision.transforms.functional_tensorrrrr#r&r)r-r.rBrErFrPrKZseterrrXrcrirxrzrrrrrrrrrrrrrrrrrrsf     # $ 1 9 6  >         0