Add basic thumbnail extraction to Rawkit (#2659)

* Add `CompressionValue`

* Rename `Transform` to `OrientationValue`

* Add basic thumbnail extraction

* Add `ThumbnailFormat`

* fmt + clippy fixes

Author: geom3trik

libraries/rawkit/src/decoder/arw1.rs

@@ -1,7 +1,7 @@
 use crate::tiff::Ifd;
 use crate::tiff::file::TiffRead;
 use crate::tiff::tags::SonyDataOffset;
-use crate::{RawImage, SubtractBlack, Transform};
+use crate::{OrientationValue, RawImage, SubtractBlack};
 use bitstream_io::{BE, BitRead, BitReader, Endianness};
 use std::io::{Read, Seek};
 
@@ -25,7 +25,7 @@ pub fn decode_a100<R: Read + Seek>(ifd: Ifd, file: &mut TiffRead<R>) -> RawImage
 		#[allow(unreachable_code)]
 		maximum: (1 << 12) - 1,
 		black: SubtractBlack::None,
-		transform: Transform::Horizontal,
+		orientation: OrientationValue::Horizontal,
 		camera_model: None,
 		camera_white_balance: None,
 		white_balance: None,

libraries/rawkit/src/decoder/arw2.rs

@@ -1,8 +1,8 @@
 use crate::tiff::file::{Endian, TiffRead};
 use crate::tiff::tags::{BitsPerSample, CfaPattern, CfaPatternDim, Compression, ImageLength, ImageWidth, SonyToneCurve, StripByteCounts, StripOffsets, Tag, WhiteBalanceRggbLevels};
-use crate::tiff::values::CurveLookupTable;
+use crate::tiff::values::{CompressionValue, CurveLookupTable};
 use crate::tiff::{Ifd, TiffError};
-use crate::{RawImage, SubtractBlack, Transform};
+use crate::{OrientationValue, RawImage, SubtractBlack};
 use rawkit_proc_macros::Tag;
 use std::io::{Read, Seek};
 
@@ -26,7 +26,7 @@ pub fn decode<R: Read + Seek>(ifd: Ifd, file: &mut TiffRead<R>) -> RawImage {
 
 	assert!(ifd.strip_offsets.len() == ifd.strip_byte_counts.len());
 	assert!(ifd.strip_offsets.len() == 1);
-	assert!(ifd.compression == 32767);
+	assert!(ifd.compression == CompressionValue::Sony_ARW_Compressed);
 
 	let image_width: usize = ifd.image_width.try_into().unwrap();
 	let image_height: usize = ifd.image_height.try_into().unwrap();
@@ -49,7 +49,7 @@ pub fn decode<R: Read + Seek>(ifd: Ifd, file: &mut TiffRead<R>) -> RawImage {
 		cfa_pattern: ifd.cfa_pattern.try_into().unwrap(),
 		maximum: (1 << 14) - 1,
 		black: SubtractBlack::CfaGrid([512, 512, 512, 512]), // TODO: Find the correct way to do this
-		transform: Transform::Horizontal,
+		orientation: OrientationValue::Horizontal,
 		camera_model: None,
 		camera_white_balance: ifd.white_balance_levels.map(|arr| arr.map(|x| x as f64)),
 		white_balance: None,

libraries/rawkit/src/decoder/uncompressed.rs

@@ -1,7 +1,8 @@
 use crate::tiff::file::TiffRead;
 use crate::tiff::tags::{BitsPerSample, BlackLevel, CfaPattern, CfaPatternDim, Compression, ImageLength, ImageWidth, RowsPerStrip, StripByteCounts, StripOffsets, Tag, WhiteBalanceRggbLevels};
+use crate::tiff::values::CompressionValue;
 use crate::tiff::{Ifd, TiffError};
-use crate::{RawImage, SubtractBlack, Transform};
+use crate::{OrientationValue, RawImage, SubtractBlack};
 use rawkit_proc_macros::Tag;
 use std::io::{Read, Seek};
 
@@ -26,7 +27,7 @@ pub fn decode<R: Read + Seek>(ifd: Ifd, file: &mut TiffRead<R>) -> RawImage {
 
 	assert!(ifd.strip_offsets.len() == ifd.strip_byte_counts.len());
 	assert!(ifd.strip_offsets.len() == 1);
-	assert!(ifd.compression == 1); // 1 is the value for uncompressed format
+	assert!(ifd.compression == CompressionValue::Uncompressed);
 
 	let image_width: usize = ifd.image_width.try_into().unwrap();
 	let image_height: usize = ifd.image_height.try_into().unwrap();
@@ -57,7 +58,7 @@ pub fn decode<R: Read + Seek>(ifd: Ifd, file: &mut TiffRead<R>) -> RawImage {
 		cfa_pattern: ifd.cfa_pattern.try_into().unwrap(),
 		maximum: if bits_per_sample == 16 { u16::MAX } else { (1 << bits_per_sample) - 1 },
 		black: SubtractBlack::CfaGrid(ifd.black_level),
-		transform: Transform::Horizontal,
+		orientation: OrientationValue::Horizontal,
 		camera_model: None,
 		camera_white_balance: ifd.white_balance_levels.map(|arr| arr.map(|x| x as f64)),
 		white_balance: None,

libraries/rawkit/src/lib.rs

@@ -12,13 +12,24 @@ 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::tags::{Compression, ImageLength, ImageWidth, Orientation, StripByteCounts, SubIfd, Tag, ThumbnailLength, ThumbnailOffset};
+use tiff::values::{CompressionValue, OrientationValue};
 use tiff::{Ifd, TiffError};
 
 pub(crate) const CHANNELS_IN_RGB: usize = 3;
 pub(crate) type Histogram = [[usize; 0x2000]; CHANNELS_IN_RGB];
 
+pub enum ThumbnailFormat {
+	Jpeg,
+	Unsupported,
+}
+
+/// A thumbnail image extracted from the raw file. This is usually a JPEG image.
+pub struct ThumbnailImage {
+	pub data: Vec<u8>,
+	pub format: ThumbnailFormat,
+}
+
 /// The amount of black level to be subtracted from Raw Image.
 pub enum SubtractBlack {
 	/// Don't subtract any value.
@@ -48,7 +59,7 @@ pub struct RawImage {
 	pub cfa_pattern: [u8; 4],
 
 	/// Transformation to be applied to negate the orientation of camera.
-	pub transform: Transform,
+	pub orientation: OrientationValue,
 
 	/// The maximum possible value of pixel that the camera sensor could give.
 	pub maximum: u16,
@@ -97,8 +108,8 @@ pub struct Image<T> {
 
 	/// The transformation required to orient the image correctly.
 	///
-	/// This will be [`Transform::Horizontal`] after the transform step is applied.
-	pub transform: Transform,
+	/// This will be [`OrientationValue::Horizontal`] after the orientation step is applied.
+	pub orientation: OrientationValue,
 }
 
 #[allow(dead_code)]
@@ -119,15 +130,15 @@ impl RawImage {
 		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 orientation = 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 {
+			if arw_ifd.compression == CompressionValue::Uncompressed {
 				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)
@@ -138,13 +149,38 @@ impl RawImage {
 		};
 
 		raw_image.camera_model = Some(camera_model);
-		raw_image.transform = transform;
+		raw_image.orientation = orientation;
 
 		raw_image.calculate_conversion_matrices();
 
 		Ok(raw_image)
 	}
 
+	/// Extracts the thumbnail image from the raw file.
+	pub fn extract_thumbnail<R: Read + Seek>(reader: &mut R) -> Result<ThumbnailImage, DecoderError> {
+		let mut file = TiffRead::new(reader)?;
+		let ifd = Ifd::new_first_ifd(&mut file)?;
+
+		// TODO: ARW files Store the thumbnail offset and length in the first IFD. Add support for other file types in the future.
+		let thumbnail_offset = ifd.get_value::<ThumbnailOffset, _>(&mut file)?;
+		let thumbnail_length = ifd.get_value::<ThumbnailLength, _>(&mut file)?;
+		file.seek_from_start(thumbnail_offset)?;
+
+		let mut thumbnail_data = vec![0; thumbnail_length as usize];
+		file.read_exact(&mut thumbnail_data)?;
+
+		// Check the first two bytes to determine the format of the thumbnail.
+		// JPEG format starts with 0xFF, 0xD8.
+		if thumbnail_data[0..2] == [0xFF, 0xD8] {
+			Ok(ThumbnailImage {
+				data: thumbnail_data,
+				format: ThumbnailFormat::Jpeg,
+			})
+		} else {
+			Err(DecoderError::UnsupportedThumbnailFormat)
+		}
+	}
+
 	/// Converts the [`RawImage`] to an [`Image`] with 8 bit resolution for each channel.
 	///
 	/// Applies all the processing steps to finally get RGB pixel data.
@@ -156,7 +192,7 @@ impl RawImage {
 			data: image.data.iter().map(|x| (x >> 8) as u8).collect(),
 			width: image.width,
 			height: image.height,
-			transform: image.transform,
+			orientation: image.orientation,
 		}
 	}
 
@@ -174,7 +210,7 @@ impl RawImage {
 		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 {
+		if image.orientation == OrientationValue::Horizontal {
 			image.apply(gamma_correction)
 		} else {
 			image.transform_and_apply(gamma_correction)
@@ -211,7 +247,7 @@ impl RawImage {
 			data: image,
 			width: self.width,
 			height: self.height,
-			transform: self.transform,
+			orientation: self.orientation,
 		}
 	}
 }
@@ -232,7 +268,7 @@ impl Image<u16> {
 
 	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();
+		let (width, height, iter) = self.orientation_iter();
 		for Pixel { values, row, column } in iter.map(|mut pixel| {
 			pixel.values = transform.apply(pixel);
 			pixel
@@ -246,7 +282,7 @@ impl Image<u16> {
 			data: image,
 			width,
 			height,
-			transform: Transform::Horizontal,
+			orientation: OrientationValue::Horizontal,
 		}
 	}
 }
@@ -259,4 +295,6 @@ pub enum DecoderError {
 	ConversionError(#[from] std::num::TryFromIntError),
 	#[error("An IO Error ocurred")]
 	IoError(#[from] std::io::Error),
+	#[error("The thumbnail format is unsupported")]
+	UnsupportedThumbnailFormat,
 }

libraries/rawkit/src/postprocessing/transform.rs

@@ -1,16 +1,16 @@
-use crate::{Image, Pixel, Transform};
+use crate::{Image, OrientationValue, Pixel};
 
 impl Image<u16> {
-	pub fn transform_iter(&self) -> (usize, usize, impl Iterator<Item = Pixel> + use<'_>) {
-		let (final_width, final_height) = if self.transform.will_swap_coordinates() {
+	pub fn orientation_iter(&self) -> (usize, usize, impl Iterator<Item = Pixel> + use<'_>) {
+		let (final_width, final_height) = if self.orientation.will_swap_coordinates() {
 			(self.height, self.width)
 		} else {
 			(self.width, self.height)
 		};
 
-		let index_0_0 = inverse_transform_index(self.transform, 0, 0, self.width, self.height);
-		let index_0_1 = inverse_transform_index(self.transform, 0, 1, self.width, self.height);
-		let index_1_0 = inverse_transform_index(self.transform, 1, 0, self.width, self.height);
+		let index_0_0 = inverse_orientation_index(self.orientation, 0, 0, self.width, self.height);
+		let index_0_1 = inverse_orientation_index(self.orientation, 0, 1, self.width, self.height);
+		let index_1_0 = inverse_orientation_index(self.orientation, 1, 0, self.width, self.height);
 
 		let column_step = (index_0_1.0 - index_0_0.0, index_0_1.1 - index_0_0.1);
 		let row_step = (index_1_0.0 - index_0_0.0, index_1_0.1 - index_0_0.1);
@@ -42,16 +42,16 @@ impl Image<u16> {
 	}
 }
 
-pub fn inverse_transform_index(transform: Transform, mut row: usize, mut column: usize, width: usize, height: usize) -> (i64, i64) {
-	let value = match transform {
-		Transform::Horizontal => 0,
-		Transform::MirrorHorizontal => 1,
-		Transform::Rotate180 => 3,
-		Transform::MirrorVertical => 2,
-		Transform::MirrorHorizontalRotate270 => 4,
-		Transform::Rotate90 => 6,
-		Transform::MirrorHorizontalRotate90 => 7,
-		Transform::Rotate270 => 5,
+pub fn inverse_orientation_index(orientation: OrientationValue, mut row: usize, mut column: usize, width: usize, height: usize) -> (i64, i64) {
+	let value = match orientation {
+		OrientationValue::Horizontal => 0,
+		OrientationValue::MirrorHorizontal => 1,
+		OrientationValue::Rotate180 => 3,
+		OrientationValue::MirrorVertical => 2,
+		OrientationValue::MirrorHorizontalRotate270 => 4,
+		OrientationValue::Rotate90 => 6,
+		OrientationValue::MirrorHorizontalRotate90 => 7,
+		OrientationValue::Rotate270 => 5,
 	};
 
 	if value & 4 != 0 {

libraries/rawkit/src/tiff/mod.rs

@@ -26,8 +26,8 @@ pub enum TagId {
 	RowsPerStrip = 0x116,
 	StripByteCounts = 0x117,
 	SubIfd = 0x14a,
-	JpegOffset = 0x201,
-	JpegLength = 0x202,
+	ThumbnailOffset = 0x201,
+	ThumbnailLength = 0x202,
 	SonyToneCurve = 0x7010,
 	BlackLevel = 0x7310,
 	WhiteBalanceRggbLevels = 0x7313,
@@ -88,10 +88,10 @@ impl Ifd {
 		}
 
 		file.seek_from_start(offset)?;
-		let num = file.read_u16()?;
+		let num_entries = file.read_u16()?;
 
-		let mut ifd_entries = Vec::with_capacity(num.into());
-		for _ in 0..num {
+		let mut ifd_entries = Vec::with_capacity(num_entries.into());
+		for _ in 0..num_entries {
 			let tag = file.read_u16()?.into();
 			let the_type = file.read_u16()?.into();
 			let count = file.read_u32()?;

libraries/rawkit/src/tiff/tags.rs

@@ -1,4 +1,4 @@
-use super::types::{Array, ConstArray, TagType, TypeByte, TypeIfd, TypeLong, TypeNumber, TypeOrientation, TypeSRational, TypeSShort, TypeShort, TypeSonyToneCurve, TypeString};
+use super::types::{Array, ConstArray, TagType, TypeByte, TypeCompression, TypeIfd, TypeLong, TypeNumber, TypeOrientation, TypeSRational, TypeSShort, TypeShort, TypeSonyToneCurve, TypeString};
 use super::{Ifd, TagId, TiffError, TiffRead};
 use std::io::{Read, Seek};
 
@@ -22,8 +22,8 @@ pub struct SamplesPerPixel;
 pub struct RowsPerStrip;
 pub struct StripByteCounts;
 pub struct SubIfd;
-pub struct JpegOffset;
-pub struct JpegLength;
+pub struct ThumbnailOffset;
+pub struct ThumbnailLength;
 pub struct SonyDataOffset;
 pub struct SonyToneCurve;
 pub struct BlackLevel;
@@ -55,7 +55,7 @@ impl SimpleTag for BitsPerSample {
 }
 
 impl SimpleTag for Compression {
-	type Type = TypeShort;
+	type Type = TypeCompression;
 
 	const ID: TagId = TagId::Compression;
 	const NAME: &'static str = "Compression";
@@ -124,17 +124,17 @@ impl SimpleTag for SubIfd {
 	const NAME: &'static str = "SubIFD";
 }
 
-impl SimpleTag for JpegOffset {
+impl SimpleTag for ThumbnailOffset {
 	type Type = TypeLong;
 
-	const ID: TagId = TagId::JpegOffset;
+	const ID: TagId = TagId::ThumbnailOffset;
 	const NAME: &'static str = "Jpeg Offset";
 }
 
-impl SimpleTag for JpegLength {
+impl SimpleTag for ThumbnailLength {
 	type Type = TypeLong;
 
-	const ID: TagId = TagId::JpegLength;
+	const ID: TagId = TagId::ThumbnailLength;
 	const NAME: &'static str = "Jpeg Length";
 }
 

libraries/rawkit/src/tiff/types.rs

@@ -1,5 +1,5 @@
 use super::file::TiffRead;
-use super::values::{CurveLookupTable, Rational, Transform};
+use super::values::{CompressionValue, CurveLookupTable, OrientationValue, Rational};
 use super::{Ifd, IfdTagType, TiffError};
 use std::io::{Read, Seek};
 
@@ -350,6 +350,7 @@ impl<T: PrimitiveType, const N: usize> TagType for ConstArray<T, N> {
 	}
 }
 
+pub struct TypeCompression;
 pub struct TypeString;
 pub struct TypeSonyToneCurve;
 pub struct TypeOrientation;
@@ -376,19 +377,17 @@ impl TagType for TypeSonyToneCurve {
 }
 
 impl TagType for TypeOrientation {
-	type Output = Transform;
+	type Output = OrientationValue;
 
 	fn read<R: Read + Seek>(file: &mut TiffRead<R>) -> Result<Self::Output, TiffError> {
-		Ok(match TypeShort::read(file)? {
-			1 => Transform::Horizontal,
-			2 => Transform::MirrorHorizontal,
-			3 => Transform::Rotate180,
-			4 => Transform::MirrorVertical,
-			5 => Transform::MirrorHorizontalRotate270,
-			6 => Transform::Rotate90,
-			7 => Transform::MirrorHorizontalRotate90,
-			8 => Transform::Rotate270,
-			_ => return Err(TiffError::InvalidValue),
-		})
+		OrientationValue::try_from(TypeShort::read(file)?).map_err(|_| TiffError::InvalidValue)
+	}
+}
+
+impl TagType for TypeCompression {
+	type Output = CompressionValue;
+
+	fn read<R: Read + Seek>(file: &mut TiffRead<R>) -> Result<Self::Output, TiffError> {
+		CompressionValue::try_from(TypeShort::read(file)?).map_err(|_| TiffError::InvalidValue)
 	}
 }