pyiqa.archs.piqe_arch

PIQE metric implementation.

Paper:

14. Venkatanath, D. Praneeth, Bh. M. Chandrasekhar, S. S. Channappayya, and S. S. Medasani. “Blind Image Quality Evaluation Using Perception Based Features”, In Proceedings of the 21st National Conference on Communications (NCC). Piscataway, NJ: IEEE, 2015.

References

• Matlab: https://www.mathworks.com/help/images/ref/piqe.html

• Python: https://github.com/michael-rutherford/pypiqe

This PyTorch implementation by: Chaofeng Chen (https://github.com/chaofengc)

Module Contents

pyiqa.archs.piqe_arch.piqe(img: torch.Tensor, block_size: int = 16, activity_threshold: float = 0.1, block_impaired_threshold: float = 0.1, window_size: int = 6) → torch.Tensor

Calculates the Perceptual Image Quality Estimator (PIQE) score for an input image. :param - img: The input image tensor. :type - img: torch.Tensor :param - block_size: The size of the blocks used for processing. Defaults to 16. :type - block_size: int, optional :param - activity_threshold: The threshold for considering a block as active. Defaults to 0.1. :type - activity_threshold: float, optional :param - block_impaired_threshold: The threshold for considering a block as impaired. Defaults to 0.1. :type - block_impaired_threshold: float, optional :param - window_size: The size of the window used for block analysis. Defaults to 6. :type - window_size: int, optional

Returns:

The PIQE score for the input image.

Return type:

• torch.Tensor

pyiqa.archs.piqe_arch.noise_criterion(block, block_size, block_var)

Function to analyze block for Gaussian noise distortions.

pyiqa.archs.piqe_arch.cal_center_sur_dev(block, block_size)

Function to compute center surround Deviation of a block.

pyiqa.archs.piqe_arch.notice_dist_criterion(blocks, window_size, block_impaired_threshold, N)

Analyze blocks for noticeable artifacts and Gaussian noise distortions.

Parameters:

• blocks (torch.Tensor) – Tensor of shape (b, num_blocks, block_size, block_size).

• window_size (int) – Size of the window for segment analysis.

• block_impaired_threshold (float) – Threshold for considering a block as impaired.

• N (int) – Size of the blocks (same as block_size).

Returns:

Tensor indicating impaired blocks.

Return type:

torch.Tensor

class pyiqa.archs.piqe_arch.PIQE

Bases: torch.nn.Module

PIQE module.

Parameters:

x (torch.Tensor) – Input tensor of shape (B, C, H, W).

Returns:

PIQE score.

Return type:

torch.Tensor

get_masks()

forward(x: torch.Tensor) → torch.Tensor