kornia-rs 0.1.5

use crate::image::Image;

/// Compute the mean squared error (MSE) between two images.
///
/// The MSE is defined as:
///
/// $ MSE = \frac{1}{n} \sum_{i=1}^{n} (I_1 - I_2)^2 $
///
/// where `I_1` and `I_2` are the two images and `n` is the number of pixels.
///
/// # Arguments
///
/// * `image1` - The first input image with shape (H, W, C).
/// * `image2` - The second input image with shape (H, W, C).
///
/// # Returns
///
/// The mean squared error between the two images.
///
/// # Example
///
/// ```
/// use kornia_rs::image::{Image, ImageSize};
///
/// let image1 = Image::<f32, 1>::new(
///    ImageSize {
///       width: 2,
///      height: 3,
/// },
/// vec![0f32, 1f32, 2f32, 3f32, 4f32, 5f32],
/// )
/// .unwrap();
///
/// let image2 = Image::<f32, 1>::new(
///    ImageSize {
///      width: 2,
///     height: 3,
/// },
/// vec![0f32, 1f32, 2f32, 3f32, 4f32, 5f32],
/// )
/// .unwrap();
///
/// let mse = kornia_rs::metrics::mse(&image1, &image2);
/// assert_eq!(mse, 0f32);
/// ```
///
/// # Panics
///
/// Panics if the two images have different shapes.
pub fn mse<const CHANNELS: usize>(
    image1: &Image<f32, CHANNELS>,
    image2: &Image<f32, CHANNELS>,
) -> f32 {
    assert_eq!(image1.size(), image2.size());
    let diff = &image1.data - &image2.data;
    diff.mapv(|x| x.powi(2)).sum() / (image1.data.len() as f32)
}

/// Compute the peak signal-to-noise ratio (PSNR) between two images.
///
/// The PSNR is defined as:
///
/// $ PSNR = 20 \log_{10} \left( \frac{MAX}{\sqrt{MSE}} \right) $
///
/// where `MAX` is the maximum possible pixel value and `MSE` is the mean squared error.
///
/// # Arguments
///
/// * `image1` - The first input image with shape (H, W, C).
/// * `image2` - The second input image with shape (H, W, C).
/// * `max_value` - The maximum possible pixel value.
///
/// # Returns
///
/// The peak signal-to-noise ratio between the two images.
///
/// # Example
/// ```
/// use kornia_rs::image::{Image, ImageSize};
///
/// let image1 = Image::<f32, 3>::new(
///   ImageSize {
///     width: 1,
///    height: 2,
/// },
/// vec![0f32, 1f32, 2f32, 3f32, 4f32, 5f32],
/// )
/// .unwrap();
///
/// let image2 = Image::<f32, 3>::new(
///   ImageSize {
///    width: 1,
///  height: 2,
/// },
/// vec![1f32, 3f32, 2f32, 4f32, 5f32, 6f32],
/// )
/// .unwrap();
///
/// let psnr = kornia_rs::metrics::psnr(&image1, &image2, 1.0);
/// assert_eq!(psnr, 320.15698);
/// ```
///
/// # Panics
///
/// Panics if the two images have different shapes.
///
/// # Note
///
/// The PSNR is expressed in decibels (dB).
///
/// The PSNR is used to measure the quality of a reconstructed image. The higher the PSNR, the better the quality of the reconstructed image.
/// The PSNR is widely used in image and video compression.
/// Underneath, the PSNR is based on the mean squared error [mse].
pub fn psnr<const CHANNELS: usize>(
    image1: &Image<f32, CHANNELS>,
    image2: &Image<f32, CHANNELS>,
    max_value: f32,
) -> f32 {
    assert_eq!(image1.size(), image2.size());
    let mse = mse(image1, image2);
    if mse == 0f32 {
        return f32::INFINITY;
    }
    20f32 * (max_value / mse.sqrt().log10())
}

#[cfg(test)]
mod tests {
    use crate::image::{Image, ImageSize};
    use anyhow::Result;

    #[test]
    fn test_equal() -> Result<()> {
        let image1 = Image::<_, 1>::new(
            ImageSize {
                width: 2,
                height: 3,
            },
            vec![0f32, 1f32, 2f32, 3f32, 4f32, 5f32],
        )?;
        let image2 = Image::<_, 1>::new(
            ImageSize {
                width: 2,
                height: 3,
            },
            vec![0f32, 1f32, 2f32, 3f32, 4f32, 5f32],
        )?;
        let mse = crate::metrics::mse(&image1, &image2);
        assert_eq!(mse, 0f32);

        Ok(())
    }

    #[test]
    fn test_not_equal() -> Result<()> {
        let image1 = Image::<_, 1>::new(
            ImageSize {
                width: 2,
                height: 2,
            },
            vec![0f32, 1f32, 2f32, 3f32],
        )?;
        let image2 = Image::<_, 1>::new(
            ImageSize {
                width: 2,
                height: 2,
            },
            vec![0f32, 3f32, 2f32, 3f32],
        )?;
        let mse = crate::metrics::mse(&image1, &image2);
        assert_eq!(mse, 1.0);

        Ok(())
    }

    #[test]
    fn test_psnr() -> Result<()> {
        let image1 = Image::<_, 3>::new(
            ImageSize {
                width: 1,
                height: 2,
            },
            vec![0f32, 1f32, 2f32, 3f32, 4f32, 5f32],
        )?;
        let image2 = Image::<_, 3>::new(
            ImageSize {
                width: 1,
                height: 2,
            },
            vec![1f32, 3f32, 2f32, 4f32, 5f32, 6f32],
        )?;
        let psnr = crate::metrics::psnr(&image1, &image2, 1.0);
        assert_eq!(psnr, 320.15698);

        Ok(())
    }
}