image-rs / image-rs/imageproc

Rotation introduces black bars

Open
#415 3 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

Dominant language
Rust
Stars
979
Forks
181
Avg merge
5m
Merged PRs (30d)
4

Description

The issue in #391 is also present in the cropping rotate function.

Rotating a square by multiples of pi/2 about centre shouldn't have any pixels with a preimage outside the original square. But calling rotate_about_center introduces bars of 99s.

I've added the following test

    fn test_rotate_half_pi_zero_square_about_center() {
        let image = gray_image!(
            00, 00, 00, 00;
            00, 00, 00, 00;
            00, 00, 00, 00;
            00, 00, 00, 00);

        let expected = gray_image!(
            00, 00, 00, 00;
            00, 00, 00, 00;
            00, 00, 00, 00;
            00, 00, 00, 00);

        let rotated = rotate_about_center(
            &image,
            std::f32::consts::PI / 2f32,
            Interpolation::Nearest,
            Luma([99u8]),
        );
        assert_pixels_eq!(rotated, expected);
    }

which fails because a bar of 99s has been introduced.

thread 'geometric_transformations::tests::test_rotate_half_pi_zero_square_about_center' panicked at 'pixels do not match.
Actual:
       0   1   2 
   +-------------
   | 
  0|  99   0   0 
   | 
  1|  99   0   0 
   | 
  2|  99   0   0 
   | 
  3|  99   0   0 
   | 
Expected:
      0  1  2 
   +----------
   | 
  0|  0  0  0 
   | 
  1|  0  0  0 
   | 
  2|  0  0  0 
   | 
  3|  0  0  0 
   | 
', src/geometric_transformations.rs:895:9

Rotating by pi introduces bars in the top and side.

Actual:
       0   1   2   3 
   +-----------------
   | 
  0|  99  99  99  99 
   | 
  1|  99   0   0   0 
   | 
  2|  99   0   0   0 
   | 
  3|  99   0   0   0 
   | 
Expected:
      0  1  2  3 
   +-------------
   | 
  0|  0  0  0  0 
   | 
  1|  0  0  0  0 
   | 
  2|  0  0  0  0 
   | 
  3|  0  0  0  0 
   | 
', src/geometric_transformations.rs:895:9

Similarly, rotating by 3pi/2 introduces a black bar at the top. Testing with non-zero squares shows that the result square is offset by one pixel.

---- geometric_transformations::tests::test_rotate_half_pi_square_about_center stdout ----
thread 'geometric_transformations::tests::test_rotate_half_pi_square_about_center' panicked at 'pixels do not match.
Actual:
       0   1   2   3 
   +-----------------
   | 
  0|  99  31  21  10 
   | 
  1|  99  32  22  11 
   | 
  2|  99  33  23  12 
   | 
  3|  99  34  25  14 
   | 
Expected:
       0   1   2   3 
   +-----------------
   | 
  0|  31  21  10   0 
   | 
  1|  32  22  11   1 
   | 
  2|  33  23  12   2 
   | 
  3|  34  25  14   3 
   | 
', src/geometric_transformations.rs:899:9

Some of these additional tests are there in this branch
https://github.com/chocolatier/imageproc/commits/fix_rotate

Contributor guide

Open the contributing guide

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

Research direction

Start in src/geometric_transformations.rs, especially rotate_about_center and the tests around line 895. Run the geometric transformation tests, including the square and zero-square cases shown in the issue, and compare the additional tests in the referenced fix_rotate branch. Done means rotations by pi/2, pi, and 3pi/2 about the center have no spurious bars or one-pixel offset.

Written by the indexing model from the issue text.

Assessment

Tech stack
rust
Domain
computer-vision
Issue type
Bug
Difficulty
4/5
Estimated time
3-5 days
Activity status
Stale
Clarity
Clearly specified
Newbie friendliness
48/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.