pyiqa.archs.brisque_arch
========================

.. py:module:: pyiqa.archs.brisque_arch

.. autoapi-nested-parse::

   BRISQUE Metric

   Created by: https://github.com/photosynthesis-team/piq/blob/master/piq/brisque.py

   Modified by: Jiadi Mo (https://github.com/JiadiMo)

   Reference:
       MATLAB codes: https://live.ece.utexas.edu/research/Quality/index_algorithms.htm BRISQUE;
       Pretrained model from: https://github.com/photosynthesis-team/piq/releases/download/v0.4.0/brisque_svm_weights.pt









Module Contents
---------------

.. py:data:: default_model_urls

.. py:function:: brisque(x: torch.Tensor, kernel_size: int = 7, kernel_sigma: float = 7 / 6, test_y_channel: bool = True, sv_coef: torch.Tensor = None, sv: torch.Tensor = None, gamma: float = 0.05, rho: float = -153.591, scale: float = 1, version: str = 'original') -> torch.Tensor

   Interface of BRISQUE index.

   :param x: An input tensor. Shape :math:`(N, C, H, W)`.
   :type x: torch.Tensor
   :param kernel_size: The side-length of the sliding window used in comparison. Must be an odd value.
   :type kernel_size: int
   :param kernel_sigma: Sigma of normal distribution.
   :type kernel_sigma: float
   :param test_y_channel: Whether to use the y-channel of YCBCR.
   :type test_y_channel: bool
   :param sv_coef: Support vector coefficients.
   :type sv_coef: torch.Tensor
   :param sv: Support vectors.
   :type sv: torch.Tensor
   :param gamma: Gamma parameter for the RBF kernel.
   :type gamma: float
   :param rho: Bias term in the decision function.
   :type rho: float
   :param scale: Scaling factor for the features.
   :type scale: float
   :param version: Version of the BRISQUE implementation ('original' or 'matlab').
   :type version: str

   :returns: Value of BRISQUE index.
   :rtype: torch.Tensor

   .. rubric:: References

   Mittal, Anish, Anush Krishna Moorthy, and Alan Conrad Bovik.
   "No-reference image quality assessment in the spatial domain."
   IEEE Transactions on image processing 21, no. 12 (2012): 4695-4708.


.. py:function:: natural_scene_statistics(luma: torch.Tensor, kernel_size: int = 7, sigma: float = 7.0 / 6) -> torch.Tensor

   Compute natural scene statistics (NSS) features for a given luminance image.

   :param luma: Luminance image tensor.
   :type luma: torch.Tensor
   :param kernel_size: Size of the Gaussian kernel.
   :type kernel_size: int
   :param sigma: Standard deviation of the Gaussian kernel.
   :type sigma: float

   :returns: NSS features.
   :rtype: torch.Tensor


.. py:function:: scale_features(features: torch.Tensor) -> torch.Tensor

   Scale features to the range [-1, 1] based on predefined feature ranges.

   :param features: Input features.
   :type features: torch.Tensor

   :returns: Scaled features.
   :rtype: torch.Tensor


.. py:function:: rbf_kernel(features: torch.Tensor, sv: torch.Tensor, gamma: float = 0.05) -> torch.Tensor

   Compute the Radial Basis Function (RBF) kernel between features and support vectors.

   :param features: Input features.
   :type features: torch.Tensor
   :param sv: Support vectors.
   :type sv: torch.Tensor
   :param gamma: Gamma parameter for the RBF kernel.
   :type gamma: float

   :returns: RBF kernel values.
   :rtype: torch.Tensor


.. py:class:: BRISQUE(kernel_size: int = 7, kernel_sigma: float = 7 / 6, test_y_channel: bool = True, version: str = 'original', pretrained_model_path: str = None)

   Bases: :py:obj:`torch.nn.Module`


   Creates a criterion that measures the BRISQUE score.

   :param kernel_size: By default, the mean and covariance of a pixel is obtained
                       by convolution with given filter_size. Must be an odd value.
   :type kernel_size: int
   :param kernel_sigma: Standard deviation for Gaussian kernel.
   :type kernel_sigma: float
   :param test_y_channel: Whether to use the y-channel of YCBCR.
   :type test_y_channel: bool
   :param version: Version of the BRISQUE implementation ('original' or 'matlab').
   :type version: str
   :param pretrained_model_path: The model path.
   :type pretrained_model_path: str, optional

   .. attribute:: kernel_size

      The side-length of the sliding window used in comparison.

      :type: int

   .. attribute:: kernel_sigma

      Sigma of normal distribution.

      :type: float

   .. attribute:: test_y_channel

      Whether to use the y-channel of YCBCR.

      :type: bool

   .. attribute:: sv_coef

      Support vector coefficients.

      :type: torch.Tensor

   .. attribute:: sv

      Support vectors.

      :type: torch.Tensor

   .. attribute:: gamma

      Gamma parameter for the RBF kernel.

      :type: float

   .. attribute:: rho

      Bias term in the decision function.

      :type: float

   .. attribute:: scale

      Scaling factor for the features.

      :type: float

   .. attribute:: version

      Version of the BRISQUE implementation ('original' or 'matlab').

      :type: str


   .. py:method:: forward(x: torch.Tensor) -> torch.Tensor

      Computation of BRISQUE score as a loss function.

      :param x: An input tensor with (N, C, H, W) shape. RGB channel order for colour images.
      :type x: torch.Tensor

      :returns: Value of BRISQUE metric.
      :rtype: torch.Tensor