feat: implement perspective warp

[EscapedGibbon]

No description provided.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 98.66%. Comparing base (798d40d) to head (f0dae9b).
Report is 1 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #484      +/-   ##
==========================================
+ Coverage   98.62%   98.66%   +0.03%     
==========================================
  Files         244      246       +2     
  Lines       10347    10645     +298     
  Branches     2141     2170      +29     
==========================================
+ Hits        10205    10503     +298     
  Misses        142      142              

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[EscapedGibbon]

Currently, there is only the test case from the issue but the implementation right now seems to be identical to the openCV implementation when nearest neighbor interpolation is used. Besides adding more test cases, should there be an option that allows the choice of interpolation?

[stropitek]

The initial idea was to implement a function which returns a transform matrix like OpenCV does, and to use that transform matrix to create the new image using Image.transform (https://image-js.github.io/image-js-typescript/classes/Image.html#transform)

The transform method already supports choosing the interpolation method.

[EscapedGibbon]

There is a function in legacy version that calculates width and height with respect of image's aspect ratio. Is it okay to remove the function that calculates width and height to preserve ratio of the image and just specify desired width and height? openCV version doesn't have one, you need to put destination corner points there. We could do a function with 4 source points and 4 destination points.

[EscapedGibbon]

Also the transform() function currently accepts 2x3 matrices for affine transformations. Should i modify the code so that it accepts only 3x3 matrices or it should accept both, and if input matrix is 2x3 i just add another row? After all, if i understand correcly, affine transformation is perspective transformation with [0,0,1] as last matrix row

[EscapedGibbon]

After discussion with @stropitek we decided to have a getPerspectiveWarp() function that returns a 3x3 transformation matrix. transform() function will be modified to accept 3x3 matrix. If 2x3 matrix is given, it will be considered an affine transformation and additional row [0,0,1] will be added to the matrix.

[stropitek]

Can you add the script which generated the images transformed by opencv to generate.py?

---

What are the new exported APIs? It does not seem that you export the new functionalities.