diff --git a/Cargo.lock b/Cargo.lock index c5d150f4..aa7ccb6b 100644 --- a/Cargo.lock +++ b/Cargo.lock @@ -1430,6 +1430,7 @@ dependencies = [ "filmcraft-codecs", "filmcraft-color", "filmcraft-frame", + "filmcraft-isobmff", "filmcraft-media", "filmcraft-time", "gif", diff --git a/crates/media/Cargo.toml b/crates/media/Cargo.toml index 19a7ba31..a3ff624b 100644 --- a/crates/media/Cargo.toml +++ b/crates/media/Cargo.toml @@ -20,6 +20,7 @@ effectcraft-pdf = { workspace = true } effectcraft-model = { workspace = true } filmcraft-codecs = { workspace = true } filmcraft-media = { workspace = true } +filmcraft-isobmff = { workspace = true } filmcraft-frame = { workspace = true } filmcraft-color = { workspace = true } filmcraft-time = { workspace = true } diff --git a/crates/media/src/convert.rs b/crates/media/src/convert.rs index f22e6e44..d483e4b4 100644 --- a/crates/media/src/convert.rs +++ b/crates/media/src/convert.rs @@ -114,6 +114,32 @@ pub(crate) fn frame_to_image_in(f: &VideoFrame, op: AlphaOp, mut buf: Vec) - img } +/// A movie frame turned the way its file says it is shown ([`crate::probe::video_turn`]: +/// clockwise quarter turns; anything else leaves it as decoded). +pub(crate) fn turned(img: Image, turn: image::metadata::Orientation) -> Image { + use image::metadata::Orientation; + let (w, h) = (img.width as usize, img.height as usize); + if img.data.len() != w * h || w == 0 || h == 0 { + return img; + } + // The source pixel shown at (x, y) of the turned picture. + let (tw, th, from): (usize, usize, fn(usize, usize, usize, usize) -> usize) = match turn { + Orientation::Rotate90 => (h, w, |x, y, w, h| (h - 1 - x) * w + y), + Orientation::Rotate180 => (w, h, |x, y, w, h| (h - 1 - y) * w + (w - 1 - x)), + Orientation::Rotate270 => (h, w, |x, y, w, _| x * w + (w - 1 - y)), + _ => return img, + }; + let mut data = vec![[0.0; 4]; tw * th]; + data.par_chunks_mut(tw).enumerate().for_each(|(y, row)| { + for (x, px) in row.iter_mut().enumerate() { + if let Some(s) = img.data.get(from(x, y, w, h)) { + *px = *s; + } + } + }); + Image { width: tw as u32, height: th as u32, data } +} + /// A FilmCraft frame for a render that shows it at `w` × `h`, at most half its size each way: /// what `resample(&frame_to_image(f, op), w, h)` makes, converting a quarter of the pixels. /// Opaque 4:2:0 Y'CbCr is box-filtered to half size before conversion (each 2×2 block of luma, diff --git a/crates/media/src/layered.rs b/crates/media/src/layered.rs index ce9bf1f9..45ef5291 100644 --- a/crates/media/src/layered.rs +++ b/crates/media/src/layered.rs @@ -87,7 +87,7 @@ pub(crate) fn decode(path: &str, bytes: &[u8], footage: &Footage, op: AlphaOp) - let inner = effectcraft_psd::Psd::parse(data.to_vec()).map_err(|e| MediaError::Decode(format!("{path} (smart object): {e}")))?; inner.composite().map_err(|e| MediaError::Decode(format!("{path} (smart object): {e}")))? } else { - let img = image::load_from_memory(data).map_err(|e| MediaError::Decode(format!("{path} (smart object): {e}")))?; + let img = crate::probe::load_oriented(data).map_err(|e| MediaError::Decode(format!("{path} (smart object): {e}")))?; let rgba = img.to_rgba32f(); effectcraft_psd::Pixels { width: rgba.width(), height: rgba.height(), data: rgba.pixels().map(|p| p.0).collect() } }; diff --git a/crates/media/src/pool.rs b/crates/media/src/pool.rs index 4ccd2463..90e19171 100644 --- a/crates/media/src/pool.rs +++ b/crates/media/src/pool.rs @@ -9,8 +9,9 @@ use effectcraft_raster::{Image, Px}; use effectcraft_render::{FloatCurve, FootageSource}; use effectcraft_time::{FrameRate, TICKS_PER_SECOND, Tick}; use filmcraft_media::{FrameRequest, SharedSource}; +use image::metadata::Orientation; -use crate::convert::{AlphaOp, dynamic_to_image, frame_to_image_in, frame_to_image_scaled}; +use crate::convert::{AlphaOp, dynamic_to_image, frame_to_image_in, frame_to_image_scaled, turned}; use crate::gif_anim::{self, Anim}; use crate::{MediaError, Result}; @@ -170,9 +171,10 @@ struct CacheState { last: HashMap, i64>, } -/// A movie/audio source by path, opened by the first caller while the others wait for it +/// A movie/audio source by path and the turn its picture is shown with +/// ([`crate::probe::video_turn`]), opened by the first caller while the others wait for it /// (`None`: failed to open; not retried until `forget`). -type SourceSlot = Arc>>; +type SourceSlot = Arc>>; /// An animated GIF by path, as [`SourceSlot`] (the error message when it failed to load). type GifSlot = Arc, String>>>; @@ -576,6 +578,11 @@ impl Inner { /// it again: a preview's frame workers, audio feeder, conform and waveform threads each /// decoded a long FLAC in full at once (#408). fn source(&self, path: &Arc) -> Option { + self.opened(path).map(|(src, _)| src) + } + + /// [`Self::source`] and the turn the movie's picture is shown with. + fn opened(&self, path: &Arc) -> Option<(SharedSource, Orientation)> { let slot = lock(&self.sources).entry(path.clone()).or_default().clone(); slot.get_or_init(|| { self.opened.fetch_add(1, Ordering::Relaxed); @@ -586,7 +593,8 @@ impl Inner { let opened = in_place.unwrap_or_else(|| { self.read(path).and_then(|bytes| { let name = std::path::Path::new(&**path).file_name().map(|n| n.to_string_lossy().to_string()).unwrap_or_default(); - filmcraft_codecs::open_bytes(&name, bytes).map_err(MediaError::from) + let turn = crate::probe::video_turn(&bytes); + filmcraft_codecs::open_bytes(&name, bytes).map(|src| (src, turn)).map_err(MediaError::from) }) }); match opened { @@ -759,7 +767,7 @@ impl Inner { } let loaded = match crate::still_format(path, &bytes) { Some(fmt) => image::load_from_memory_with_format(&bytes, fmt), - None => image::load_from_memory(&bytes), + None => crate::probe::load_oriented(&bytes), }; let img = match loaded { Ok(img) => img, @@ -774,19 +782,22 @@ impl Inner { Ok(dynamic_to_image(frame, op)) } Some(mt) => { - let src = self.source(path).ok_or_else(|| MediaError::Io(format!("{path}: cannot open")))?; + let (src, turn) = self.opened(path).ok_or_else(|| MediaError::Io(format!("{path}: cannot open")))?; let vf = src.video_frame(FrameRequest::full(filmcraft_time::Tick(mt.0))).map_err(MediaError::from)?; + // A movie shown turned a quarter (a phone held upright): the footage's size and + // the render's are the turned picture's, the decoded frame's are the file's. + let side = |(w, h): (u32, u32)| if crate::probe::swaps_sides(turn) { (h, w) } else { (w, h) }; // A reduced-resolution render's frame, made at its size (a frame of the footage's // own size only, as the renderer resamples only those). - let (w, h) = loc.key.size; + let (w, h) = side(loc.key.size); if w > 0 - && (vf.width, vf.height) == (footage.width, footage.height) + && (vf.width, vf.height) == side((footage.width, footage.height)) && let Some(img) = frame_to_image_scaled(&vf, w, h) { - return Ok(img); + return Ok(turned(img, turn)); } let buf = lock(&self.cache).lru.take_spare(vf.width as usize * vf.height as usize); - Ok(frame_to_image_in(&vf, op, buf)) + Ok(turned(frame_to_image_in(&vf, op, buf), turn)) } } } diff --git a/crates/media/src/probe.rs b/crates/media/src/probe.rs index d8ea5506..a27f0209 100644 --- a/crates/media/src/probe.rs +++ b/crates/media/src/probe.rs @@ -5,6 +5,7 @@ use std::sync::Arc; use effectcraft_project::{AlphaMode, Footage, FootageKind}; use effectcraft_time::{FrameRate, Tick}; +use image::metadata::Orientation; use crate::{MediaError, Result, ext_of, rate_from_fc}; @@ -90,8 +91,17 @@ pub fn probe_bytes(path: &str, bytes: Arc<[u8]>) -> Result { return probe_still_bytes(path, &bytes, fmt); } let name = Path::new(path).file_name().map(|n| n.to_string_lossy().to_string()).unwrap_or_else(|| path.to_string()); + let turn = video_turn(&bytes); let src = filmcraft_codecs::open_bytes(&name, bytes)?; - Ok(footage_from_info(path, src.info())) + let mut f = footage_from_info(path, src.info()); + // The size as displayed: a movie shot with the camera on its side trades width and height. + if swaps_sides(turn) { + std::mem::swap(&mut f.width, &mut f.height); + if f.pixel_aspect > 0.0 { + f.pixel_aspect = 1.0 / f.pixel_aspect; + } + } + Ok(f) } pub(crate) fn footage_from_info(path: &str, info: &filmcraft_media::MediaInfo) -> Footage { @@ -181,7 +191,12 @@ fn probe_still_bytes(path: &str, bytes: &[u8], fmt: image::ImageFormat) -> Resul } let reader = image::ImageReader::with_format(std::io::Cursor::new(bytes), fmt); let (w, h, has_alpha) = match reader.into_decoder() { - Ok(dec) => (dec.dimensions().0, dec.dimensions().1, dec.color_type().has_alpha()), + Ok(mut dec) => { + // The size as displayed: an Exif orientation of a quarter turn swaps the sides. + let (w, h) = dec.dimensions(); + let (w, h) = if swaps_sides(dec.orientation().unwrap_or(Orientation::NoTransforms)) { (h, w) } else { (w, h) }; + (w, h, dec.color_type().has_alpha()) + } Err(e) => { // A multi-layer OpenEXR file without an unnamed RGB layer. let img = (fmt == image::ImageFormat::OpenExr).then(|| crate::exr_channels::layered_image(bytes)).flatten(); @@ -191,3 +206,39 @@ fn probe_still_bytes(path: &str, bytes: &[u8], fmt: image::ImageFormat) -> Resul }; Ok(still_footage(path, w, h, fmt, has_alpha)) } + +/// Whether an Exif orientation turns the picture a quarter turn (its width and height trade places). +pub(crate) fn swaps_sides(o: Orientation) -> bool { + matches!(o, Orientation::Rotate90 | Orientation::Rotate270 | Orientation::Rotate90FlipH | Orientation::Rotate270FlipH) +} + +/// Decode a still the way it is meant to be seen: its Exif orientation (a camera held sideways, a +/// mirrored scan) is applied to the pixels, as image viewers do. A file without +/// the tag, or with a broken one, decodes as stored. +pub(crate) fn load_oriented(bytes: &[u8]) -> image::ImageResult { + use image::ImageDecoder; + let mut dec = image::ImageReader::new(std::io::Cursor::new(bytes)).with_guessed_format()?.into_decoder()?; + let orientation = dec.orientation().unwrap_or(Orientation::NoTransforms); + let mut img = image::DynamicImage::from_decoder(dec)?; + img.apply_orientation(orientation); + Ok(img) +} + +/// The rotation an MP4/MOV movie is shown with (a phone held upright records sideways frames and +/// says so in its video track's header): the quarter turn nearest the track's rotation. Only the +/// rotation counts; any mirroring, scaling or offset in the header is ignored, and other +/// containers have none. +pub(crate) fn video_turn(bytes: &[u8]) -> Orientation { + let Ok(file) = filmcraft_isobmff::open(bytes) else { return Orientation::NoTransforms }; + let Some(m) = file.tracks.iter().find(|t| t.kind == filmcraft_isobmff::TrackKind::Video).map(|t| t.matrix) else { + return Orientation::NoTransforms; + }; + // [a b u; c d v; x y w]: (a, b) is where the picture's x axis points. + let (a, b) = (m[0] as i64, m[1] as i64); + match (a.abs() >= b.abs(), a >= 0, b >= 0) { + (true, true, _) => Orientation::NoTransforms, + (true, false, _) => Orientation::Rotate180, + (false, _, true) => Orientation::Rotate90, + (false, _, false) => Orientation::Rotate270, + } +} diff --git a/crates/media/tests/exif_orientation.rs b/crates/media/tests/exif_orientation.rs new file mode 100644 index 00000000..d5acc3c2 --- /dev/null +++ b/crates/media/tests/exif_orientation.rs @@ -0,0 +1,84 @@ +//! A still's Exif orientation (a camera held sideways, a mirrored scan) is applied on import: the +//! footage has the size the picture is seen at, and its frames are the picture the right way up. + +use std::sync::Arc; + +use effectcraft_media::{MediaPool, probe_bytes}; +use effectcraft_project::FootageKind; +use effectcraft_time::Tick; + +const RED: [f32; 3] = [1.0, 0.0, 0.0]; +const GREEN: [f32; 3] = [0.0, 1.0, 0.0]; +const BLUE: [f32; 3] = [0.0, 0.0, 1.0]; +const WHITE: [f32; 3] = [1.0, 1.0, 1.0]; + +/// A 32×16 JPEG with red, green / blue, white quarters, tagged with Exif orientation `tag` +/// (`None`: no Exif at all). +fn quartered_jpeg(tag: Option) -> Vec { + let img = image::RgbImage::from_fn(32, 16, |x, y| { + let c = match (x < 16, y < 8) { + (true, true) => RED, + (false, true) => GREEN, + (true, false) => BLUE, + (false, false) => WHITE, + }; + image::Rgb(c.map(|v| (v * 255.0) as u8)) + }); + let mut jpeg = Vec::new(); + image::codecs::jpeg::JpegEncoder::new_with_quality(&mut jpeg, 100).encode_image(&img).expect("encode"); + let Some(tag) = tag else { return jpeg }; + // APP1 "Exif": a big-endian TIFF header and one IFD holding Orientation (0x0112, SHORT). + let mut app1 = vec![0xFF, 0xE1, 0x00, 0x22]; + app1.extend_from_slice(b"Exif\0\0MM\0\x2A\0\0\0\x08"); + app1.extend_from_slice(&[0x00, 0x01, 0x01, 0x12, 0x00, 0x03, 0x00, 0x00, 0x00, 0x01, 0x00, tag, 0x00, 0x00]); + app1.extend_from_slice(&[0x00; 4]); + // (Straight after the SOI marker.) + jpeg.splice(2..2, app1); + jpeg +} + +/// The footage size and the colours of its four corners (top-left, top-right, bottom-left, +/// bottom-right), as imported. +fn imported(tag: Option) -> ((u32, u32), [[f32; 3]; 4]) { + let name = format!("oriented-{tag:?}.jpg"); + let bytes: Arc<[u8]> = quartered_jpeg(tag).into(); + let f = probe_bytes(&name, bytes.clone()).expect("probe"); + assert_eq!(f.kind, FootageKind::Still); + let pool = MediaPool::new(); + pool.add_bytes(&name, bytes); + let img = pool.frame_at(&f, Tick::ZERO).expect("frame"); + assert_eq!((img.width, img.height), (f.width, f.height), "orientation {tag:?}: the frame has the probed size"); + let (w, h) = (img.width as usize, img.height as usize); + let at = |x: usize, y: usize| { + let p = img.data[y * w + x]; + [p[0], p[1], p[2]] + }; + ((f.width, f.height), [at(2, 2), at(w - 3, 2), at(2, h - 3), at(w - 3, h - 3)]) +} + +#[test] +fn exif_orientation_is_applied_to_stills() { + let (landscape, portrait) = ((32, 16), (16, 32)); + let cases = [ + (None, landscape, [RED, GREEN, BLUE, WHITE]), + (Some(1), landscape, [RED, GREEN, BLUE, WHITE]), + (Some(2), landscape, [GREEN, RED, WHITE, BLUE]), + (Some(3), landscape, [WHITE, BLUE, GREEN, RED]), + (Some(4), landscape, [BLUE, WHITE, RED, GREEN]), + (Some(5), portrait, [RED, BLUE, GREEN, WHITE]), + (Some(6), portrait, [BLUE, RED, WHITE, GREEN]), + (Some(7), portrait, [WHITE, GREEN, BLUE, RED]), + (Some(8), portrait, [GREEN, WHITE, RED, BLUE]), + // (Not an orientation: the picture as stored.) + (Some(0), landscape, [RED, GREEN, BLUE, WHITE]), + (Some(9), landscape, [RED, GREEN, BLUE, WHITE]), + ]; + for (tag, size, corners) in cases { + let (got_size, got) = imported(tag); + assert_eq!(got_size, size, "orientation {tag:?}: size"); + for (i, (g, want)) in got.iter().zip(corners).enumerate() { + let near = g.iter().zip(want).all(|(a, b)| (a - b).abs() < 0.2); + assert!(near, "orientation {tag:?}: corner {i} is {g:?}, expected {want:?}"); + } + } +} diff --git a/crates/media/tests/footage.rs b/crates/media/tests/footage.rs index 1c0a97dd..4734eb35 100644 --- a/crates/media/tests/footage.rs +++ b/crates/media/tests/footage.rs @@ -9,7 +9,7 @@ use std::process::Command; use std::sync::Arc; use std::time::Instant; -use effectcraft_media::{MediaPool, probe}; +use effectcraft_media::{MediaPool, probe, probe_bytes}; use effectcraft_project::{AlphaMode, Footage, FootageKind, ItemId}; use effectcraft_raster::Image; use effectcraft_render::FootageSource; @@ -336,6 +336,59 @@ fn resampled_audio_read_in_frame_pieces_matches_ffmpeg() { } } +/// The H.264 fixture with the rotation part of its video track header's matrix (`[a b; c d]`, +/// 16.16 fixed point) replaced: what a phone writes for a movie shot with the camera turned. +fn turned_mp4(p: &Path, abcd: [i32; 4]) -> Arc<[u8]> { + let mut bytes = std::fs::read(p).expect("read"); + // (The video track comes first; a version 0 `tkhd` has its matrix 44 bytes after the type.) + let at = bytes.windows(4).position(|w| w == b"tkhd").expect("tkhd"); + assert_eq!(bytes[at + 4], 0, "a version 0 tkhd"); + for (slot, v) in [0, 1, 3, 4].into_iter().zip(abcd) { + let o = at + 44 + slot * 4; + bytes[o..o + 4].copy_from_slice(&v.to_be_bytes()); + } + bytes.into() +} + +/// A movie whose track header says it is shown rotated (a phone held upright) imports the way it +/// is shown: the footage has the turned size and its frames are turned, at full and reduced size. +#[test] +fn rotated_mp4_imports_upright() { + let Some(p) = h264_mp4() else { return }; + let plain_footage = probe(&p).expect("probe"); + let pool = MediaPool::new(); + let t = t_of(plain_footage.frame_rate, 3); + let plain = pool.frame(ItemId(1), &plain_footage, t).expect("frame"); + let (w, h) = (plain.width as usize, plain.height as usize); + let one = 0x1_0000; + // The matrix, the turned size and the plain frame's pixel shown at (x, y). + type From = fn(usize, usize, usize, usize) -> (usize, usize); + let cases: [(&str, [i32; 4], (usize, usize), From); 3] = [ + ("90", [0, one, -one, 0], (h, w), |x, y, _, h| (y, h - 1 - x)), + ("180", [-one, 0, 0, -one], (w, h), |x, y, w, h| (w - 1 - x, h - 1 - y)), + ("270", [0, -one, one, 0], (h, w), |x, y, w, _| (w - 1 - y, x)), + ]; + for (name, matrix, (tw, th), from) in cases { + let path = format!("turned-{name}.mp4"); + let bytes = turned_mp4(&p, matrix); + let f = probe_bytes(&path, bytes.clone()).expect("probe"); + assert_eq!((f.width as usize, f.height as usize), (tw, th), "{name}°: size"); + pool.add_bytes(&path, bytes); + let img = pool.frame(ItemId(2), &f, t).expect("frame"); + assert_eq!((img.width as usize, img.height as usize), (tw, th), "{name}°: frame size"); + for (x, y) in [(0, 0), (tw - 1, 0), (0, th - 1), (tw - 1, th - 1), (tw / 3, th / 5), (tw / 2, th - 7)] { + let (sx, sy) = from(x, y, w, h); + assert_eq!(img.data[y * tw + x], plain.data[sy * w + sx], "{name}°: pixel ({x}, {y})"); + } + // (A reduced-resolution render asks for the turned size.) + let (hw, hh) = (f.width / 2, f.height / 2); + let sized = pool.frame_at_size(ItemId(2), &f, t, hw, hh).expect("frame"); + assert_eq!((sized.width, sized.height), (hw, hh), "{name}°: reduced size"); + let db = psnr(&sized, &effectcraft_raster::resample(&img, hw, hh)); + assert!(db > 45.0, "{name}° at half size: {db:.2} dB"); + } +} + #[test] fn prores_4444_with_alpha() { let Some(p) = prores_mov() else { return };