The Stitching Pipeline can be extended by estimate cropping ROIs in low resolution:

Therefore, the warped low resolution images can be blended together to create a blend mask on which the largest interior rectangle (lir) can be found. At the moment the outline approach of the lir repository is used, which can detect the lir in 0.1s in a 1 MP image. The Algorithm used is Marzeh, Tahmasbi and Mireh (2019): Algorithm for finding the largest inscribed rectangle in polygon, optimized on the outline of a binary raster. On the long run, it might be worth to integrate the functionality natively in opencv. Using the lir (red), the cropping roi (blue) of a single image (green) can be determined and used for cropping both the small and the full resolution images.

Cropping the images on low resolution and before estimating the seam masks has mayor benefits:

1. The Estimation is faster
2. Without cropping the seams need to go from image corner to image corner:

With cropping the seam carving is much more natural:

I already updated the opencv_stitching_tutorial with a more detailed explaination of the cropping procedure.

Pull Request Readiness Checklist

See details at https://github.com/opencv/opencv/wiki/How_to_contribute#making-a-good-pull-request

• I agree to contribute to the project under Apache 2 License.
• To the best of my knowledge, the proposed patch is not based on a code under GPL or other license that is incompatible with OpenCV
• The PR is proposed to proper branch
• There is reference to original bug report and related work
• There is accuracy test, performance test and test data in opencv_extra repository, if applicable
  Patch to opencv_extra has the same branch name.
• The feature is well documented and sample code can be built with the project CMake

force_builders=linux,docs