[OpenVINO] Implement image transforms (affine, perspective, elastic)

[goyaladitya05]

Description

This PR implements affine_transform, perspective_transform, and elastic_transform for the OpenVINO backend.

• affine_transform: builds transformed coords via OV range/broadcast/matmul, right-multiplies pts @ T + offset per the scipy convention, then delegates to map_coordinates
• perspective_transform: solves the 8×8 system via ov_opset.inverse, applies the projective warp formula per output pixel
• elastic_transform: generates random displacement fields with Box-Muller sampling, smooths with gaussian_blur, then displaces coords before sampling

Removed 50 tests, which now pass after these implementations.

Closes: openvinotoolkit/openvino/issues/35412

Contributor Agreement

Please check all boxes below before submitting your PR for review:

• I am a human, and not a bot.
• I will be responsible for responding to review comments in a timely manner.
• I will work with the maintainers to push this PR forward until submission.

Note: Failing to adhere to this agreement may result in your future PRs no longer being reviewed.

[gemini-code-assist]

Code Review

This pull request implements affine_transform, perspective_transform, and elastic_transform for the OpenVINO backend, enabling several previously excluded image operation tests. The implementation uses OpenVINO opsets for coordinate mapping and homography computation. Feedback identifies a hardcoded fill_mode in perspective_transform that should be parameterized for API consistency, a docstring error regarding return dimensions, and opportunities to optimize coordinate gathering. Additionally, the review suggests improving robustness against division-by-zero in projective transformations and notes a performance bottleneck in elastic_transform due to a symbolic-to-NumPy conversion during seeding.

[codecov-commenter]

Codecov Report

❌ Patch coverage is 92.77108% with 24 lines in your changes missing coverage. Please review.
✅ Project coverage is 77.68%. Comparing base (e5c9a6e) to head (82088c6).
⚠️ Report is 2 commits behind head on master.

Files with missing lines	Patch %	Lines
keras/src/backend/openvino/image.py	92.77%	12 Missing and 12 partials ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##           master   #22696      +/-   ##
==========================================
- Coverage   78.82%   77.68%   -1.15%     
==========================================
  Files         596      596              
  Lines       69293    69625     +332     
  Branches    10829    10861      +32     
==========================================
- Hits        54622    54089     -533     
- Misses      11941    12781     +840     
- Partials     2730     2755      +25     

Flag	Coverage Δ
keras	77.59% <92.77%> (-1.10%)	⬇️
keras-jax	58.43% <2.10%> (?)
keras-numpy	54.28% <0.00%> (-0.26%)	⬇️
keras-openvino	59.68% <92.77%> (+0.25%)	⬆️
keras-tensorflow	59.98% <2.10%> (-0.28%)	⬇️
keras-torch	?

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ 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.

[hertschuh]

Thanks for implementing these!