fovea 0.5.0

A high-precision, type-safe computer vision library guaranteeing absolute image correctness at compile time
Documentation
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
//! Two-pass union-find connected-components engine.
//!
//! Implements the two-pass union-find engine. See the module-level docs
//! for the public surface and the worked 4×4 example.

use crate::image::{Image, ImageView, ImageViewMut, RasterImage};
use crate::pixel::LabelPixel;
use crate::{Coordinate, Error};

use super::Labeling;
use super::connectivity::Connectivity;
use super::measurements::BlobMeasurements;
use super::stats::ComponentStats;
use super::stats::sink::{NoStats, StatsSink, WithMeasurements, WithStats};
use super::union_find::UnionFind;

/// Compute the connected-component labeling of `image`, allocating a
/// fresh [`Labeling<L>`].
///
/// The label pixel type `L` and connectivity strategy `C` are named
/// explicitly by the caller (turbofish), e.g.
/// `connected_components::<Label32, Connectivity8>(&binary)`.
///
/// # Errors — Tier 2
///
/// Returns [`Error::LabelOverflow`] if the input contains more
/// connected components than `L::MAX_LABEL` can encode. The count is of
/// final components, not provisional labels, so an image whose components
/// merge through many provisional labels is not rejected. Also returned,
/// with `label_capacity == u32::MAX`, when the engine's own `u32`
/// provisional-label space is exhausted, which takes more than
/// `u32::MAX` disconnected foreground sites and is out of practical
/// reach.
///
/// # Examples
///
/// ```
/// use fovea::analyze::components::{connected_components, Connectivity4};
/// use fovea::image::BinaryImage;
/// use fovea::pixel::Label32;
///
/// let img = BinaryImage::fill(8, 8, true);
/// let r = connected_components::<Label32, Connectivity4>(&img).unwrap();
/// assert_eq!(r.label_count, 1);
/// ```
pub fn connected_components<L, C>(
    image: &impl RasterImage<Pixel = bool>,
) -> Result<Labeling<L>, Error>
where
    L: LabelPixel,
    C: Connectivity,
{
    let mut labels = Image::<L>::zero(image.width(), image.height());
    let label_count = connected_components_into::<L, C>(image, &mut labels)?;
    Ok(Labeling {
        labels,
        label_count,
    })
}

/// Compute the connected-component labeling of `image`, writing the
/// label image into `out` and returning `label_count`.
///
/// On an `Err` return the contents of `out` are unspecified: the
/// capacity check runs while labels are being written, so a prefix of
/// the buffer may already carry labels from the failed run.
///
/// # Panics
///
/// Panics if `out.size() != image.size()` (Tier 3 — programmer bug).
///
/// # Errors \u2014 Tier 2
///
/// Returns [`Error::LabelOverflow`] if the input contains more
/// connected components than `L::MAX_LABEL` can encode. The count is of
/// final components, not provisional labels, so an image whose components
/// merge through many provisional labels is not rejected. Also returned,
/// with `label_capacity == u32::MAX`, when the engine's own `u32`
/// provisional-label space is exhausted, which takes more than
/// `u32::MAX` disconnected foreground sites and is out of practical
/// reach.
pub fn connected_components_into<L, C>(
    image: &impl RasterImage<Pixel = bool>,
    out: &mut Image<L>,
) -> Result<u32, Error>
where
    L: LabelPixel,
    C: Connectivity,
{
    assert_eq!(
        out.size(),
        image.size(),
        "connected_components_into: output size {:?} does not match input {:?}",
        out.size(),
        image.size()
    );

    run::<L, C, _, NoStats>(image, out, &mut NoStats)
}

/// Compute the connected-component labeling of `image`, plus one
/// [`ComponentStats`] per foreground component (area, bounding box,
/// centroid sums).
///
/// Stats are accumulated inline during pass 2 so the image is not
/// rescanned after labeling.
///
/// # Errors \u2014 Tier 2
///
/// Returns [`Error::LabelOverflow`] if the input contains more
/// connected components than `L::MAX_LABEL` can encode. The count is of
/// final components, not provisional labels, so an image whose components
/// merge through many provisional labels is not rejected. Also returned,
/// with `label_capacity == u32::MAX`, when the engine's own `u32`
/// provisional-label space is exhausted, which takes more than
/// `u32::MAX` disconnected foreground sites and is out of practical
/// reach.
///
/// # Examples
///
/// ```
/// use fovea::analyze::components::{
///     connected_components_with_stats, Connectivity4,
/// };
/// use fovea::image::BinaryImage;
/// use fovea::pixel::Label32;
///
/// // A 2x2 square.
/// let img = BinaryImage::fill(2, 2, true);
/// let (lab, stats) =
///     connected_components_with_stats::<Label32, Connectivity4>(&img).unwrap();
/// assert_eq!(lab.label_count, 1);
/// assert_eq!(stats[0].area, 4);
/// assert_eq!(stats[0].centroid(), fovea::CoordinateF64::new(0.5, 0.5));
/// ```
pub fn connected_components_with_stats<L, C>(
    image: &impl RasterImage<Pixel = bool>,
) -> Result<(Labeling<L>, Vec<ComponentStats>), Error>
where
    L: LabelPixel,
    C: Connectivity,
{
    let mut labels = Image::<L>::zero(image.width(), image.height());
    let mut stats: Vec<ComponentStats> = Vec::new();
    let label_count = {
        let mut sink = WithStats { out: &mut stats };
        run::<L, C, _, WithStats<'_>>(image, &mut labels, &mut sink)?
    };
    debug_assert_eq!(stats.len(), label_count as usize);
    Ok((
        Labeling {
            labels,
            label_count,
        },
        stats,
    ))
}

/// Compute the connected-component labeling of `image`, plus one
/// [`BlobMeasurements`] per foreground component: area, bounding box,
/// centroid sums, the raw second-order moment sums, and a 4-connected
/// boundary-pixel count. From these the shape descriptors
/// (equivalent diameter, orientation, eccentricity, circularity) are
/// derived on demand — see [`BlobMeasurements`].
///
/// Everything is accumulated in the *same* single pass 2 as the labeling;
/// there is no separate contour extraction. This is the heavier sibling
/// of [`connected_components_with_stats`]: it additionally runs a
/// per-foreground-pixel 4-neighbour boundary check to count the
/// boundary pixels. Reach for [`connected_components_with_stats`] when
/// you only need area / bounding box / centroid.
///
/// # Two non-obvious contracts
///
/// - **The boundary test is 4-connected regardless of the
///   labeling [`Connectivity`] `C`.** `C` decides which pixels form a
///   blob; the boundary test decides how its outline is counted. A
///   `Connectivity8` caller still gets a 4-connected boundary count —
///   correct, because the boundary is a property of the blob's pixel set,
///   not of the grouping rule.
/// - **Measurements are view-relative.** A blob clipped by the view edge
///   is measured as clipped: its cut edge counts toward the boundary and
///   `area` / `bbox` / centroid cover only the in-view part. When tiling,
///   use an overlapping margin and keep only blobs whose full extent lies
///   in the non-overlapped core.
///
/// # Errors — Tier 2
///
/// Returns [`Error::LabelOverflow`] if the input contains more connected
/// components than `L::MAX_LABEL` can encode. The count is of final
/// components, not provisional labels, so an image whose components merge
/// through many provisional labels is not rejected. Also returned, with
/// `label_capacity == u32::MAX`, when the engine's own `u32`
/// provisional-label space is exhausted, which takes more than
/// `u32::MAX` disconnected foreground sites and is out of practical
/// reach.
///
/// # Examples
///
/// ```
/// use fovea::analyze::components::{
///     connected_components_with_measurements, Connectivity4,
/// };
/// use fovea::image::BinaryImage;
/// use fovea::pixel::Label32;
///
/// // A solid 3x3 square: area 9, boundary pixels 8 (4·3 − 4).
/// let img = BinaryImage::fill(3, 3, true);
/// let (lab, m) =
///     connected_components_with_measurements::<Label32, Connectivity4>(&img).unwrap();
/// assert_eq!(lab.label_count, 1);
/// assert_eq!(m[0].area, 9);
/// assert_eq!(m[0].boundary_pixels, 8);
/// ```
pub fn connected_components_with_measurements<L, C>(
    image: &impl RasterImage<Pixel = bool>,
) -> Result<(Labeling<L>, Vec<BlobMeasurements>), Error>
where
    L: LabelPixel,
    C: Connectivity,
{
    let mut labels = Image::<L>::zero(image.width(), image.height());
    let mut measurements: Vec<BlobMeasurements> = Vec::new();
    let label_count = {
        let mut sink = WithMeasurements {
            out: &mut measurements,
        };
        run::<L, C, _, WithMeasurements<'_>>(image, &mut labels, &mut sink)?
    };
    debug_assert_eq!(measurements.len(), label_count as usize);
    Ok((
        Labeling {
            labels,
            label_count,
        },
        measurements,
    ))
}

// ──────────────────────────────────────────────────────────────────────
// Engine \u2014 monomorphised over `S: StatsSink`.
// ──────────────────────────────────────────────────────────────────────

/// Maximum number of raster-preceding neighbours examined per pixel
/// across every shipped [`Connectivity`]. Currently 4 (for
/// [`Connectivity8`](super::Connectivity8)). The pass-1 inner buffer
/// `others: [u32; MAX_NEIGHBOURS]` hardcodes this constant. Adding a
/// connectivity with more predecessors requires lifting this and is
/// flagged for follow-up design.
const MAX_NEIGHBOURS: usize = 4;

fn run<L, C, I, S>(image: &I, out: &mut Image<L>, sink: &mut S) -> Result<u32, Error>
where
    L: LabelPixel,
    C: Connectivity,
    I: RasterImage<Pixel = bool>,
    S: StatsSink,
{
    let w = image.width();
    let h = image.height();
    if w == 0 || h == 0 {
        return Ok(0);
    }

    // ── Pass 1 ───────────────────────────────────────────────────────
    // Provisional labels live in a flat W*H Vec<u32>, raster-scan
    // order. Zero is the background sentinel. `u32` is the width of
    // `LabelPixel::MAX_LABEL`, and this buffer is read and written by
    // both full-image passes, so its width is the engine's memory
    // traffic.
    let mut prov: Vec<u32> = vec![0; w * h];
    // Capacity hint: pathological all-stripes input produces ~W*H/4
    // labels; using that as the initial allocation keeps `make_set`
    // amortised cheap without over-allocating in the common case.
    let cap_hint = (w * h) / 4 + 1;
    let mut uf = UnionFind::with_capacity(cap_hint);

    for y in 0..h {
        let row = image.row(y);
        for x in 0..w {
            if !row[x] {
                continue;
            }

            // Collect provisional labels of the already-visited
            // foreground neighbours. The smallest is tracked
            // separately; everything else goes in `others`, which is
            // unioned with `smallest` at the end. `0` doubles as the
            // "none seen yet" sentinel: provisional labels start at 1,
            // so no foreground neighbour can carry it (`u32::MAX` could
            // not serve — it is itself a valid label).
            let mut smallest: u32 = 0;
            let mut others: [u32; MAX_NEIGHBOURS] = [0; MAX_NEIGHBOURS];
            let mut other_count = 0usize;

            let site = Coordinate::new(x, y);
            for &offset in C::OFFSETS {
                let Some(n) = site.checked_add(offset) else {
                    continue;
                };
                if n.x >= w || n.y >= h {
                    continue;
                }
                let p = prov[n.y * w + n.x];
                if p == 0 {
                    continue;
                }
                if smallest == 0 {
                    smallest = p;
                } else if p < smallest {
                    others[other_count] = smallest;
                    other_count += 1;
                    smallest = p;
                } else if p != smallest {
                    others[other_count] = p;
                    other_count += 1;
                }
            }

            let label = if smallest == 0 {
                // `make_set` returns `None` only when the u32 label
                // space itself is spent: more than `u32::MAX` provisional
                // labels, the engine's own width. The capacity of the
                // target type `L` is deliberately NOT checked here:
                // provisional labels routinely exceed the final component
                // count (merging them is what pass 2 is for), so `L` is
                // checked in pass 2 against the compacted count.
                match uf.make_set() {
                    Some(new_label) => new_label,
                    None => {
                        return Err(Error::LabelOverflow {
                            label_capacity: u32::MAX,
                        });
                    }
                }
            } else {
                for &o in &others[..other_count] {
                    uf.union(smallest, o);
                }
                smallest
            };

            prov[y * w + x] = label;
        }
    }

    // ── Pass 2 ───────────────────────────────────────────────────────
    // Resolve roots and compact labels to a dense `1..=label_count`,
    // writing the output pixels and forwarding `(label, first, x, y)`
    // to the stats sink.
    let mut compact: Vec<u32> = vec![0; uf.len()];
    // Compact labels assigned so far. `label_count + 1` cannot wrap:
    // a new root only appears while `label_count` is strictly below
    // the provisional-label total, which pass 1 capped at `u32::MAX`.
    let mut label_count: u32 = 0;

    for y in 0..h {
        for x in 0..w {
            let p = prov[y * w + x];
            let cell = out.pixel_at_mut(x, y);
            if p == 0 {
                *cell = L::zero();
            } else {
                let root = uf.find(p);
                let existing = compact[root as usize];
                let (c, first) = if existing == 0 {
                    // The `L` capacity check lives here, on the compacted
                    // count, because this is the first point at which the
                    // *final* number of components is known. Checking the
                    // provisional counter in pass 1 would reject images
                    // whose components merely merge through many labels.
                    if label_count == L::MAX_LABEL {
                        return Err(Error::LabelOverflow {
                            label_capacity: L::MAX_LABEL,
                        });
                    }
                    let assigned = label_count + 1;
                    compact[root as usize] = assigned;
                    label_count = assigned;
                    (assigned, true)
                } else {
                    (existing, false)
                };
                // Invariant: 0 < c <= label_count <= L::MAX_LABEL
                // (the capacity check just above guarantees this).
                debug_assert!(
                    c <= L::MAX_LABEL,
                    "internal invariant violated: compact label {} > MAX_LABEL {}",
                    c,
                    L::MAX_LABEL
                );
                *cell = L::from_label_index(c).expect(
                    "internal error: compact label exceeds L::MAX_LABEL despite \
                     the pass-2 capacity check (analyze::components engine)",
                );
                // The boundary neighbour-check is gated behind the sink's
                // `NEEDS_BOUNDARY` const so it is const-folded away (and
                // `is_boundary` reduces to a literal `false`) for the
                // `NoStats` / `WithStats` paths.
                let is_boundary = if S::NEEDS_BOUNDARY {
                    is_4_boundary(image, x, y)
                } else {
                    false
                };
                sink.record(c, first, Coordinate::new(x, y), is_boundary);
            }
        }
    }

    Ok(label_count)
}

/// Returns `true` if the foreground pixel at `(x, y)` is a 4-connected
/// boundary pixel: at least one of its four orthogonal neighbours is
/// background (`false`) or lies off the analyzed view.
///
/// The caller guarantees `(x, y)` is itself foreground. Off-view
/// neighbours (`image.get` → `None`) count as boundary — this is what
/// makes measurements *view-relative*: a blob clipped by the view edge
/// has its cut edge counted as boundary.
///
/// This test is fixed at 4-connectivity regardless of the labeling
/// [`Connectivity`], because a blob's boundary is a property of its
/// pixel *set*, not of the rule that grouped the pixels.
#[inline]
fn is_4_boundary<I>(image: &I, x: usize, y: usize) -> bool
where
    I: RasterImage<Pixel = bool>,
{
    // Left, right, up, down. `checked_sub` handles the x==0 / y==0 edges
    // (underflow → off-view → boundary); `get` handles the far edges.
    let left = x.checked_sub(1).and_then(|nx| image.get(nx, y));
    let right = image.get(x + 1, y);
    let up = y.checked_sub(1).and_then(|ny| image.get(x, ny));
    let down = image.get(x, y + 1);
    // A neighbour that is off-view (`None`) or background (`Some(false)`)
    // makes this a boundary pixel.
    [left, right, up, down]
        .iter()
        .any(|n| !matches!(n, Some(true)))
}

// ══════════════════════════════════════════════════════════════════════
// Tests
// ══════════════════════════════════════════════════════════════════════

#[cfg(test)]
mod tests {
    use super::super::{
        Connectivity4, Connectivity8, Labeling, connected_components, connected_components_into,
        connected_components_with_stats,
    };
    use super::connected_components_with_measurements;
    use crate::Error;
    use crate::image::{BinaryImage, Image, ImageView, SubView};
    use crate::pixel::{Label32, LabelPixel, ZeroablePixel};
    use crate::{Coordinate, Rectangle, Size};

    // Helpers ─────────────────────────────────────────────────────────

    /// Build a binary image from a string where `#` is foreground and
    /// any other non-whitespace char is background. Lines must all
    /// have the same width.
    fn img_from_str(text: &str) -> BinaryImage {
        let lines: Vec<&str> = text
            .lines()
            .map(|l| l.trim())
            .filter(|l| !l.is_empty())
            .collect();
        let h = lines.len();
        let w = lines[0].chars().count();
        for l in &lines {
            assert_eq!(l.chars().count(), w);
        }
        let mut data = Vec::with_capacity(w * h);
        for l in &lines {
            for c in l.chars() {
                data.push(c == '#');
            }
        }
        BinaryImage::from_vec(w, h, data).unwrap()
    }

    /// Group `(x, y)` foreground pixel positions by their compact
    /// label. Returns a vector of label-coordinate sets, indexed by
    /// `compact_label - 1`, plus the foreground-count total.
    fn partition_by_label(
        lab: &Labeling<Label32>,
    ) -> Vec<std::collections::BTreeSet<(usize, usize)>> {
        use std::collections::BTreeSet;
        let mut groups: Vec<BTreeSet<(usize, usize)>> =
            (0..lab.label_count).map(|_| BTreeSet::new()).collect();
        for y in 0..lab.labels.height() {
            for x in 0..lab.labels.width() {
                let v = lab.labels.pixel_at(x, y).value();
                if v != 0 {
                    groups[(v - 1) as usize].insert((x, y));
                }
            }
        }
        groups
    }

    // Engine tests ────────────────────────────────────────────────────

    #[test]
    fn empty_image_returns_zero_labels() {
        let img = BinaryImage::from_vec(0, 0, Vec::new()).unwrap();
        let r = connected_components::<Label32, Connectivity4>(&img).unwrap();
        assert_eq!(r.label_count, 0);
        assert_eq!(r.labels.size(), Size::new(0, 0));
    }

    #[test]
    fn all_background() {
        let img = BinaryImage::fill(8, 8, false);
        let r = connected_components::<Label32, Connectivity4>(&img).unwrap();
        assert_eq!(r.label_count, 0);
        for y in 0..8 {
            for x in 0..8 {
                assert_eq!(r.labels.pixel_at(x, y), Label32::BACKGROUND);
            }
        }
    }

    #[test]
    fn all_foreground_4connected() {
        let img = BinaryImage::fill(5, 4, true);
        let r = connected_components::<Label32, Connectivity4>(&img).unwrap();
        assert_eq!(r.label_count, 1);
        for y in 0..4 {
            for x in 0..5 {
                assert_eq!(r.labels.pixel_at(x, y), Label32::new(1));
            }
        }
    }

    #[test]
    fn all_foreground_8connected() {
        let img = BinaryImage::fill(5, 4, true);
        let r = connected_components::<Label32, Connectivity8>(&img).unwrap();
        assert_eq!(r.label_count, 1);
    }

    #[test]
    fn worked_4x4_example_conn4() {
        let img = img_from_str(
            r#"
            .##.
            ##..
            ..##
            ..#.
        "#,
        );
        let r = connected_components::<Label32, Connectivity4>(&img).unwrap();
        assert_eq!(r.label_count, 2);
        // The partition: 4-pixel top L and 3-pixel bottom L.
        let groups = partition_by_label(&r);
        let sizes: Vec<usize> = groups.iter().map(|g| g.len()).collect();
        let mut sorted = sizes.clone();
        sorted.sort();
        assert_eq!(sorted, vec![3, 4]);
        // Total foreground = 7.
        let total: usize = sizes.iter().sum();
        assert_eq!(total, 7);
    }

    #[test]
    fn worked_4x4_example_conn8_merges_diagonal() {
        let img = img_from_str(
            r#"
            .##.
            ##..
            ..##
            ..#.
        "#,
        );
        let r = connected_components::<Label32, Connectivity8>(&img).unwrap();
        // The diagonal at (2,2) touches (1,1) so all 8 foreground pixels collapse.
        assert_eq!(r.label_count, 1);
    }

    #[test]
    fn u_shape_forces_pass1_merge_conn4() {
        let img = img_from_str(
            r#"
            #.#
            #.#
            ###
        "#,
        );
        let r = connected_components::<Label32, Connectivity4>(&img).unwrap();
        assert_eq!(r.label_count, 1);
    }

    #[test]
    fn single_pixel_blob_at_corner() {
        let mut data = vec![false; 4 * 4];
        data[0] = true;
        let img = BinaryImage::from_vec(4, 4, data).unwrap();
        let r = connected_components::<Label32, Connectivity4>(&img).unwrap();
        assert_eq!(r.label_count, 1);
        assert_eq!(r.labels.pixel_at(0, 0), Label32::new(1));
    }

    #[test]
    fn single_row() {
        let img = BinaryImage::from_vec(5, 1, vec![true, false, true, true, false]).unwrap();
        let r = connected_components::<Label32, Connectivity4>(&img).unwrap();
        assert_eq!(r.label_count, 2);
    }

    #[test]
    fn single_column() {
        let img = BinaryImage::from_vec(1, 5, vec![true, false, true, true, false]).unwrap();
        let r = connected_components::<Label32, Connectivity4>(&img).unwrap();
        assert_eq!(r.label_count, 2);
    }

    #[test]
    fn checkerboard_conn4_yields_one_per_pixel() {
        // 8x8 checkerboard, true at (x+y)%2==0
        let mut data = Vec::with_capacity(64);
        for y in 0..8 {
            for x in 0..8 {
                data.push((x + y) % 2 == 0);
            }
        }
        let img = BinaryImage::from_vec(8, 8, data).unwrap();
        let r = connected_components::<Label32, Connectivity4>(&img).unwrap();
        assert_eq!(r.label_count, 32);
    }

    #[test]
    fn checkerboard_conn8_yields_one_component() {
        let mut data = Vec::with_capacity(64);
        for y in 0..8 {
            for x in 0..8 {
                data.push((x + y) % 2 == 0);
            }
        }
        let img = BinaryImage::from_vec(8, 8, data).unwrap();
        let r = connected_components::<Label32, Connectivity8>(&img).unwrap();
        assert_eq!(r.label_count, 1);
    }

    #[test]
    fn subview_input_round_trips() {
        // Outer 6x6 image; ROI is the inner 4x4.
        let img = img_from_str(
            r#"
            ......
            .####.
            .#..#.
            .#..#.
            .####.
            ......
        "#,
        );
        let roi = img
            .roi(Rectangle::new(Coordinate::new(1, 1), Size::new(4, 4)))
            .unwrap();
        let r = connected_components::<Label32, Connectivity4>(&roi).unwrap();
        // The ring around the 4x4 ROI is a single component (12 pixels).
        assert_eq!(r.label_count, 1);
        let groups = partition_by_label(&r);
        assert_eq!(groups[0].len(), 12);
    }

    #[test]
    #[should_panic(expected = "does not match input")]
    fn into_size_mismatch_panics() {
        let img = BinaryImage::fill(4, 4, false);
        let mut out: Image<Label32> = Image::zero(8, 8);
        let _ = connected_components_into::<Label32, Connectivity4>(&img, &mut out);
    }

    #[test]
    fn label_count_matches_between_entry_points() {
        let img = img_from_str(
            r#"
            #.#.#
            .....
            #.#.#
            .....
            #.#.#
        "#,
        );
        let r = connected_components::<Label32, Connectivity4>(&img).unwrap();
        let mut out: Image<Label32> = Image::zero(5, 5);
        let count = connected_components_into::<Label32, Connectivity4>(&img, &mut out).unwrap();
        assert_eq!(count, r.label_count);
        // And the label images agree pixel-for-pixel.
        for y in 0..5 {
            for x in 0..5 {
                assert_eq!(out.pixel_at(x, y), r.labels.pixel_at(x, y));
            }
        }
    }

    #[test]
    fn total_foreground_area_matches_input() {
        let img = img_from_str(
            r#"
            ##..##
            .#..#.
            ..##..
            ##..##
            ##..##
        "#,
        );
        let r = connected_components::<Label32, Connectivity4>(&img).unwrap();
        // Count input foreground pixels.
        let mut fg = 0usize;
        for y in 0..img.height() {
            for x in 0..img.width() {
                if img.pixel_at(x, y) {
                    fg += 1;
                }
            }
        }
        // Sum component areas from stats.
        let (_, stats) = connected_components_with_stats::<Label32, Connectivity4>(&img).unwrap();
        let total_area: u64 = stats.iter().map(|s| s.area).sum();
        assert_eq!(total_area as usize, fg);
        assert_eq!(stats.len(), r.label_count as usize);
    }

    // Stats tests ─────────────────────────────────────────────────────

    #[test]
    fn stats_bbox_is_tight() {
        // Single 3x2 rectangle at (2..=4, 1..=2).
        let img = img_from_str(
            r#"
            .......
            ..###..
            ..###..
            .......
        "#,
        );
        let (_, stats) = connected_components_with_stats::<Label32, Connectivity4>(&img).unwrap();
        assert_eq!(stats.len(), 1);
        assert_eq!(stats[0].area, 6);
        assert_eq!(stats[0].bbox_min, Coordinate::new(2, 1));
        assert_eq!(stats[0].bbox_max_inclusive, Coordinate::new(4, 2));
        let bb = stats[0].bbox();
        assert_eq!(bb, Rectangle::new(Coordinate::new(2, 1), Size::new(3, 2)));
    }

    #[test]
    fn stats_centroid_of_centred_square() {
        // 3x3 square at (1..=3, 1..=3) in a 5x5 image
        let img = img_from_str(
            r#"
            .....
            .###.
            .###.
            .###.
            .....
        "#,
        );
        let (_, stats) = connected_components_with_stats::<Label32, Connectivity4>(&img).unwrap();
        assert_eq!(stats.len(), 1);
        let c = stats[0].centroid();
        assert!((c.x - 2.0).abs() < 1e-9);
        assert!((c.y - 2.0).abs() < 1e-9);
    }

    #[test]
    fn stats_multi_component() {
        // Three components: 1x1 dot at (0,0); 2x2 square at (3..=4,
        // 0..=1); 1x3 vertical bar at (0..=0, 3..=5).
        let img = img_from_str(
            r#"
            #..##.
            ...##.
            ......
            #.....
            #.....
            #.....
        "#,
        );
        let (lab, stats) = connected_components_with_stats::<Label32, Connectivity4>(&img).unwrap();
        assert_eq!(lab.label_count, 3);
        assert_eq!(stats.len(), 3);
        // Find each component by area.
        let mut sorted = stats.clone();
        sorted.sort_by_key(|s| s.area);
        // 1x1 dot
        assert_eq!(sorted[0].area, 1);
        // 1x3 bar
        assert_eq!(sorted[1].area, 3);
        assert_eq!(
            sorted[1].bbox(),
            Rectangle::new(Coordinate::new(0, 3), Size::new(1, 3))
        );
        // 2x2 square
        assert_eq!(sorted[2].area, 4);
        assert_eq!(
            sorted[2].bbox(),
            Rectangle::new(Coordinate::new(3, 0), Size::new(2, 2))
        );
    }

    // Overflow test using a test-only narrow `LabelPixel`. ────────────

    /// Test-only label type with `MAX_LABEL = 3`. Wraps a `u8`.
    #[derive(Clone, Copy, Debug, PartialEq, Eq, Hash, Ord, PartialOrd)]
    struct TinyLabel(u8);

    impl ZeroablePixel for TinyLabel {
        fn zero() -> Self {
            TinyLabel(0)
        }
    }

    impl LabelPixel for TinyLabel {
        const MAX_LABEL: u32 = 3;
        fn from_label_index(i: u32) -> Option<Self> {
            if i == 0 || i > 3 {
                None
            } else {
                Some(TinyLabel(i as u8))
            }
        }
        fn to_label_index(self) -> u32 {
            self.0 as u32
        }
    }

    #[test]
    fn label_overflow_when_components_exceed_capacity() {
        // Five horizontally-isolated dots in a single row \u2192 5
        // components, but TinyLabel::MAX_LABEL == 3.
        let img = BinaryImage::from_vec(
            9,
            1,
            vec![true, false, true, false, true, false, true, false, true],
        )
        .unwrap();
        // First sanity-check with Label32 that there really are 5 components.
        let r = connected_components::<Label32, Connectivity4>(&img).unwrap();
        assert_eq!(r.label_count, 5);

        let err = connected_components::<TinyLabel, Connectivity4>(&img).unwrap_err();
        match err {
            Error::LabelOverflow { label_capacity } => assert_eq!(label_capacity, 3),
            other => panic!("expected LabelOverflow, got {:?}", other),
        }
    }

    #[test]
    fn label_overflow_succeeds_when_at_capacity() {
        // Exactly 3 components \u2014 should fit TinyLabel.
        let img = BinaryImage::from_vec(5, 1, vec![true, false, true, false, true]).unwrap();
        let mut out: Image<TinyLabel> = Image::zero(5, 1);
        let n = connected_components_into::<TinyLabel, Connectivity4>(&img, &mut out).unwrap();
        assert_eq!(n, 3);
        assert_eq!(out.pixel_at(0, 0), TinyLabel(1));
        assert_eq!(out.pixel_at(2, 0), TinyLabel(2));
        assert_eq!(out.pixel_at(4, 0), TinyLabel(3));
    }

    #[test]
    fn label_overflow_counts_final_components_not_provisional_labels() {
        // Row 0 alone creates four provisional labels, all merged into
        // one component by row 1. TinyLabel holds 3 labels, so a check
        // against the provisional counter would reject this ordinary
        // image; the documented contract counts final components (one).
        let img = img_from_str(
            r#"
            #.#.#.#
            #######
        "#,
        );
        let r = connected_components::<Label32, Connectivity4>(&img).unwrap();
        assert_eq!(r.label_count, 1);

        let mut out: Image<TinyLabel> = Image::zero(7, 2);
        let n = connected_components_into::<TinyLabel, Connectivity4>(&img, &mut out).unwrap();
        assert_eq!(n, 1);
    }

    #[test]
    fn label_overflow_still_fires_on_merged_components_over_capacity() {
        // Eight provisional labels merging down to four components, one
        // more than TinyLabel can hold: still an overflow, reported with
        // the label type's capacity.
        let img = img_from_str(
            r#"
            #.#.#.#.#.#.#.#
            ###.###.###.###
        "#,
        );
        let r = connected_components::<Label32, Connectivity4>(&img).unwrap();
        assert_eq!(r.label_count, 4);

        let err = connected_components::<TinyLabel, Connectivity4>(&img).unwrap_err();
        match err {
            Error::LabelOverflow { label_capacity } => assert_eq!(label_capacity, 3),
            other => panic!("expected LabelOverflow, got {:?}", other),
        }
    }

    // Step 9 \u2014 trait audit on Labeling ────────────────────────────────

    #[test]
    fn labeling_is_clone_and_debug() {
        let img = BinaryImage::fill(2, 2, true);
        let r = connected_components::<Label32, Connectivity4>(&img).unwrap();
        let cloned: Labeling<Label32> = r.clone();
        assert_eq!(cloned.label_count, 1);
        let s = format!("{:?}", cloned);
        assert!(s.contains("Labeling"));
    }

    // Measurements entry-point tests ──────────────────────────────────

    #[test]
    fn measurements_boundary_pixels_of_square_is_exact() {
        // Solid 5x5 square → boundary-pixel count = 4·5 − 4 = 16.
        let img = BinaryImage::fill(5, 5, true);
        let (lab, m) =
            connected_components_with_measurements::<Label32, Connectivity4>(&img).unwrap();
        assert_eq!(lab.label_count, 1);
        assert_eq!(m.len(), 1);
        assert_eq!(m[0].area, 25);
        assert_eq!(m[0].boundary_pixels, 16);
        // Moments match the cheap path for the shared fields.
        assert_eq!(m[0].centroid(), crate::CoordinateF64::new(2.0, 2.0));
    }

    #[test]
    fn measurements_single_pixel_blob_is_finite() {
        let mut data = vec![false; 3 * 3];
        data[4] = true; // centre pixel
        let img = BinaryImage::from_vec(3, 3, data).unwrap();
        let (_, m) =
            connected_components_with_measurements::<Label32, Connectivity4>(&img).unwrap();
        assert_eq!(m.len(), 1);
        assert_eq!(m[0].area, 1);
        assert_eq!(m[0].boundary_pixels, 1);
        assert_eq!(m[0].eccentricity(), 0.0);
        assert!(m[0].orientation().radians().is_finite());
        assert!(m[0].circularity().is_finite());
    }

    #[test]
    fn measurements_moments_match_cheap_path() {
        // The shared fields (area, bbox, sum_x/y) must agree with
        // connected_components_with_stats — regression guard for the sink
        // signature change.
        let img = img_from_str(
            r#"
            #..##.
            ...##.
            ......
            #.....
            #.....
            #.....
        "#,
        );
        let (_, stats) = connected_components_with_stats::<Label32, Connectivity4>(&img).unwrap();
        let (_, meas) =
            connected_components_with_measurements::<Label32, Connectivity4>(&img).unwrap();
        assert_eq!(stats.len(), meas.len());
        for (s, m) in stats.iter().zip(meas.iter()) {
            assert_eq!(s.area, m.area);
            assert_eq!(s.bbox_min, m.bbox_min);
            assert_eq!(s.bbox_max_inclusive, m.bbox_max_inclusive);
            assert_eq!(s.sum_x, m.sum_x);
            assert_eq!(s.sum_y, m.sum_y);
        }
    }

    #[test]
    fn measurements_connectivity4_vs_8_boundary_pixels() {
        // Two pixels touching only diagonally. Under Connectivity4 they
        // are two separate 1-pixel blobs (boundary count 1 each); under
        // Connectivity8 they form one blob whose boundary count is the
        // 4-connected boundary count of the two-pixel set = 2 (each pixel
        // has a background 4-neighbour, so both are boundary pixels). The
        // boundary test stays 4-connected regardless of the labeling C.
        let img = img_from_str(
            r#"
            #..
            .#.
            ...
        "#,
        );
        let (lab4, m4) =
            connected_components_with_measurements::<Label32, Connectivity4>(&img).unwrap();
        assert_eq!(lab4.label_count, 2);
        assert_eq!(m4.len(), 2);
        for m in &m4 {
            assert_eq!(m.area, 1);
            assert_eq!(m.boundary_pixels, 1);
        }

        let (lab8, m8) =
            connected_components_with_measurements::<Label32, Connectivity8>(&img).unwrap();
        assert_eq!(lab8.label_count, 1);
        assert_eq!(m8.len(), 1);
        assert_eq!(m8[0].area, 2);
        assert_eq!(m8[0].boundary_pixels, 2);
    }

    #[test]
    fn measurements_roi_clips_boundary_pixels() {
        // A solid 3-wide, full-height bar in a 5x5 image; the ROI is the
        // left 3x3 corner. Inside the ROI the visible blob is a 3x3 solid
        // square, but its right and bottom edges are cut by the view, so
        // those pixels still count as boundary (off-view neighbour). The
        // clipped square measures area 9, boundary count 8 — as if it were a
        // standalone 3x3 square — confirming the view-relative contract.
        let img = img_from_str(
            r#"
            #####
            #####
            #####
            #####
            #####
        "#,
        );
        let roi = img
            .roi(Rectangle::new(Coordinate::new(0, 0), Size::new(3, 3)))
            .unwrap();
        let (lab, m) =
            connected_components_with_measurements::<Label32, Connectivity4>(&roi).unwrap();
        assert_eq!(lab.label_count, 1);
        assert_eq!(m[0].area, 9);
        // 3x3 block with all four view edges cutting it → every pixel is a
        // boundary pixel except the centre → boundary count 8.
        assert_eq!(m[0].boundary_pixels, 8);
        assert_eq!(
            m[0].bbox(),
            Rectangle::new(Coordinate::new(0, 0), Size::new(3, 3))
        );
    }
}