o oi=@sddlmZddlmZmZmZddlmZmZm Z ddl m Z m Z ddl mZdedeedefd d Z ddededededededefddZGdddeZdS))List)ModuleTensorpad) KORNIA_CHECKKORNIA_CHECK_IS_TENSORKORNIA_CHECK_SHAPE)filter3dget_gaussian_kernel3d)_compute_paddingimgcropping_shapereturnc Cs:t||d |d |d |d |d |d fS)z.Crop out the part of "valid" convolution area.r)r)r r rI/home/ubuntu/.local/lib/python3.10/site-packages/kornia/metrics/ssim3d.py_cropsr?-q=sameimg1img2 window_sizemax_valepspaddingcCst|t|t|gdt|gdt|j|jkd|jd|jt|ts5tdt|t|||fd}d|d}d|d}t ||} t ||} g} |d kru|jd d \} } }t | | |g} t | | } t | | } n|d krz | d}| d}| | }t |d|}t |d|}t |||}|d krt || }t || }t || }n|d kr ||}||}||}d ||d ||}||||||}|||S)abCompute the Structural Similarity (SSIM) index map between two images. Measures the (SSIM) index between each element in the input `x` and target `y`. The index can be described as: .. math:: \text{SSIM}(x, y) = \frac{(2\mu_x\mu_y+c_1)(2\sigma_{xy}+c_2)} {(\mu_x^2+\mu_y^2+c_1)(\sigma_x^2+\sigma_y^2+c_2)} where: - :math:`c_1=(k_1 L)^2` and :math:`c_2=(k_2 L)^2` are two variables to stabilize the division with weak denominator. - :math:`L` is the dynamic range of the pixel-values (typically this is :math:`2^{\#\text{bits per pixel}}-1`). Args: img1: the first input image with shape :math:`(B, C, D, H, W)`. img2: the second input image with shape :math:`(B, C, D, H, W)`. window_size: the size of the gaussian kernel to smooth the images. max_val: the dynamic range of the images. eps: Small value for numerically stability when dividing. padding: ``'same'`` | ``'valid'``. Whether to only use the "valid" convolution area to compute SSIM to match the MATLAB implementation of original SSIM paper. Returns: The ssim index map with shape :math:`(B, C, D, H, W)`. Examples: >>> input1 = torch.rand(1, 4, 5, 5, 5) >>> input2 = torch.rand(1, 4, 5, 5, 5) >>> ssim_map = ssim3d(input1, input2, 5) # 1x4x5x5x5 )BCDHWz,img1 and img2 shapes must be the same. Got: z and z'Input max_val type is not a float. Got )?r%r%g{Gz?rgQ?validNrg@) rrrshape isinstancefloat TypeErrortyper r r r)rrrrrrkernelC1C2mu1mu2r depthheightwidthmu1_sqmu2_sqmu1_mu2 mu_img1_sq mu_img2_sq mu_img1_img2 sigma1_sq sigma2_sqsigma12numdenrrrssim3d)sL&"           r@c sNeZdZdZddedededed d f fd d Zd eded efddZ Z S)SSIM3Da=Create a module that computes the Structural Similarity (SSIM) index between two 3D images. Measures the (SSIM) index between each element in the input `x` and target `y`. The index can be described as: .. math:: \text{SSIM}(x, y) = \frac{(2\mu_x\mu_y+c_1)(2\sigma_{xy}+c_2)} {(\mu_x^2+\mu_y^2+c_1)(\sigma_x^2+\sigma_y^2+c_2)} where: - :math:`c_1=(k_1 L)^2` and :math:`c_2=(k_2 L)^2` are two variables to stabilize the division with weak denominator. - :math:`L` is the dynamic range of the pixel-values (typically this is :math:`2^{\#\text{bits per pixel}}-1`). Args: window_size: the size of the gaussian kernel to smooth the images. max_val: the dynamic range of the images. eps: Small value for numerically stability when dividing. padding: ``'same'`` | ``'valid'``. Whether to only use the "valid" convolution area to compute SSIM to match the MATLAB implementation of original SSIM paper. Shape: - Input: :math:`(B, C, D, H, W)`. - Target :math:`(B, C, D, H, W)`. - Output: :math:`(B, C, D, H, W)`. Examples: >>> input1 = torch.rand(1, 4, 5, 5, 5) >>> input2 = torch.rand(1, 4, 5, 5, 5) >>> ssim = SSIM3D(5) >>> ssim_map = ssim(input1, input2) # 1x4x5x5x5 rrrrrrrrNcs&t||_||_||_||_dSN)super__init__rrrr)selfrrrr __class__rrrDs  zSSIM3D.__init__rrcCst|||j|j|j|jSrB)r@rrrr)rErrrrrforwardszSSIM3D.forwardrrr) __name__ __module__ __qualname____doc__intr*strrDrrH __classcell__rrrFrrAs$%rANrI)typingr kornia.corerrrkornia.core.checkrrrkornia.filtersr r kornia.filters.filterr rNrr*rOr@rArrrrs.   ^