openfree's picture
Deploy from GitHub repository
2409829 verified
pub mod decoder;
pub mod demosaicing;
pub mod metadata;
pub mod postprocessing;
pub mod preprocessing;
pub mod processing;
pub mod tiff;
use crate::metadata::identify::CameraModel;
use processing::{Pixel, PixelTransform, RawPixel, RawPixelTransform};
use rawkit_proc_macros::Tag;
use std::io::{Read, Seek};
use thiserror::Error;
use tiff::file::TiffRead;
use tiff::tags::{Compression, ImageLength, ImageWidth, Orientation, StripByteCounts, SubIfd, Tag};
use tiff::values::Transform;
use tiff::{Ifd, TiffError};
pub(crate) const CHANNELS_IN_RGB: usize = 3;
pub(crate) type Histogram = [[usize; 0x2000]; CHANNELS_IN_RGB];
/// The amount of black level to be subtracted from Raw Image.
pub enum SubtractBlack {
/// Don't subtract any value.
None,
/// Subtract a singular value for all pixels in Bayer CFA Grid.
Value(u16),
/// Subtract the appropriate value for pixels in Bayer CFA Grid.
CfaGrid([u16; 4]),
}
/// Represents a Raw Image along with its metadata.
pub struct RawImage {
/// Raw pixel data stored in linear fashion.
pub data: Vec<u16>,
/// Width of the raw image.
pub width: usize,
/// Height of the raw image.
pub height: usize,
/// Bayer CFA pattern used to arrange pixels in [`RawImage::data`].
///
/// It encodes Red, Blue and Green as 0, 1, and 2 respectively.
pub cfa_pattern: [u8; 4],
/// Transformation to be applied to negate the orientation of camera.
pub transform: Transform,
/// The maximum possible value of pixel that the camera sensor could give.
pub maximum: u16,
/// The minimum possible value of pixel that the camera sensor could give.
///
/// Used to subtract the black level from the raw image.
pub black: SubtractBlack,
/// Information regarding the company and model of the camera.
pub camera_model: Option<CameraModel>,
/// White balance specified in the metadata of the raw file.
///
/// It represents the 4 values of CFA Grid which follows the same pattern as [`RawImage::cfa_pattern`].
pub camera_white_balance: Option<[f64; 4]>,
/// White balance of the raw image.
///
/// It is the same as [`RawImage::camera_white_balance`] if the raw file contains the metadata.
/// Otherwise it falls back to calculating the white balance from the color space conversion matrix.
///
/// It represents the 4 values of CFA Grid which follows the same pattern as [`RawImage::cfa_pattern`].
pub white_balance: Option<[f64; 4]>,
/// Color space conversion matrix to convert from camera's color space to sRGB.
pub camera_to_rgb: Option<[[f64; 3]; 3]>,
}
/// Represents the final RGB Image.
pub struct Image<T> {
/// Pixel data stored in a linear fashion.
pub data: Vec<T>,
/// Width of the image.
pub width: usize,
/// Height of the image.
pub height: usize,
/// The number of color channels in the image.
///
/// We can assume this will be 3 for all non-obscure, modern cameras.
/// See <https://github.com/GraphiteEditor/Graphite/pull/1923#discussion_r1725070342> for more information.
pub channels: u8,
/// The transformation required to orient the image correctly.
///
/// This will be [`Transform::Horizontal`] after the transform step is applied.
pub transform: Transform,
}
#[allow(dead_code)]
#[derive(Tag)]
struct ArwIfd {
image_width: ImageWidth,
image_height: ImageLength,
compression: Compression,
strip_byte_counts: StripByteCounts,
}
impl RawImage {
/// Create a [`RawImage`] from an input stream.
///
/// Decodes the contents of `reader` and extracts raw pixel data and metadata.
pub fn decode<R: Read + Seek>(reader: &mut R) -> Result<RawImage, DecoderError> {
let mut file = TiffRead::new(reader)?;
let ifd = Ifd::new_first_ifd(&mut file)?;
let camera_model = metadata::identify::identify_camera_model(&ifd, &mut file).unwrap();
let transform = ifd.get_value::<Orientation, _>(&mut file)?;
let mut raw_image = if camera_model.model == "DSLR-A100" {
decoder::arw1::decode_a100(ifd, &mut file)
} else {
let sub_ifd = ifd.get_value::<SubIfd, _>(&mut file)?;
let arw_ifd = sub_ifd.get_value::<ArwIfd, _>(&mut file)?;
if arw_ifd.compression == 1 {
decoder::uncompressed::decode(sub_ifd, &mut file)
} else if arw_ifd.strip_byte_counts[0] == arw_ifd.image_width * arw_ifd.image_height {
decoder::arw2::decode(sub_ifd, &mut file)
} else {
// TODO: implement for arw 1.
todo!()
}
};
raw_image.camera_model = Some(camera_model);
raw_image.transform = transform;
raw_image.calculate_conversion_matrices();
Ok(raw_image)
}
/// Converts the [`RawImage`] to an [`Image`] with 8 bit resolution for each channel.
///
/// Applies all the processing steps to finally get RGB pixel data.
pub fn process_8bit(self) -> Image<u8> {
let image = self.process_16bit();
Image {
channels: image.channels,
data: image.data.iter().map(|x| (x >> 8) as u8).collect(),
width: image.width,
height: image.height,
transform: image.transform,
}
}
/// Converts the [`RawImage`] to an [`Image`] with 16 bit resolution for each channel.
///
/// Applies all the processing steps to finally get RGB pixel data.
pub fn process_16bit(self) -> Image<u16> {
let subtract_black = self.subtract_black_fn();
let scale_white_balance = self.scale_white_balance_fn();
let scale_to_16bit = self.scale_to_16bit_fn();
let raw_image = self.apply((subtract_black, scale_white_balance, scale_to_16bit));
let convert_to_rgb = raw_image.convert_to_rgb_fn();
let mut record_histogram = raw_image.record_histogram_fn();
let image = raw_image.demosaic_and_apply((convert_to_rgb, &mut record_histogram));
let gamma_correction = image.gamma_correction_fn(&record_histogram.histogram);
if image.transform == Transform::Horizontal {
image.apply(gamma_correction)
} else {
image.transform_and_apply(gamma_correction)
}
}
}
impl RawImage {
pub fn apply(mut self, mut transform: impl RawPixelTransform) -> RawImage {
for (index, value) in self.data.iter_mut().enumerate() {
let pixel = RawPixel {
value: *value,
row: index / self.width,
column: index % self.width,
};
*value = transform.apply(pixel);
}
self
}
pub fn demosaic_and_apply(self, mut transform: impl PixelTransform) -> Image<u16> {
let mut image = vec![0; self.width * self.height * 3];
for Pixel { values, row, column } in self.linear_demosaic_iter().map(|mut pixel| {
pixel.values = transform.apply(pixel);
pixel
}) {
let pixel_index = row * self.width + column;
image[3 * pixel_index..3 * (pixel_index + 1)].copy_from_slice(&values);
}
Image {
channels: 3,
data: image,
width: self.width,
height: self.height,
transform: self.transform,
}
}
}
impl Image<u16> {
pub fn apply(mut self, mut transform: impl PixelTransform) -> Image<u16> {
for (index, values) in self.data.chunks_exact_mut(3).enumerate() {
let pixel = Pixel {
values: values.try_into().unwrap(),
row: index / self.width,
column: index % self.width,
};
values.copy_from_slice(&transform.apply(pixel));
}
self
}
pub fn transform_and_apply(self, mut transform: impl PixelTransform) -> Image<u16> {
let mut image = vec![0; self.width * self.height * 3];
let (width, height, iter) = self.transform_iter();
for Pixel { values, row, column } in iter.map(|mut pixel| {
pixel.values = transform.apply(pixel);
pixel
}) {
let pixel_index = row * width + column;
image[3 * pixel_index..3 * (pixel_index + 1)].copy_from_slice(&values);
}
Image {
channels: 3,
data: image,
width,
height,
transform: Transform::Horizontal,
}
}
}
#[derive(Error, Debug)]
pub enum DecoderError {
#[error("An error occurred when trying to parse the TIFF format")]
TiffError(#[from] TiffError),
#[error("An error occurred when converting integer from one type to another")]
ConversionError(#[from] std::num::TryFromIntError),
#[error("An IO Error ocurred")]
IoError(#[from] std::io::Error),
}