I am looking for someone who can write some code to extract the rectangle card out of a complex image. Attached are a couple of the more difficult items.
I am assuming that someone has already done something like this and code reuse is welcome.
------------------ 5/16 - Adding more detailed requirements: ------------------
This project is to extract cards (rectangles) from their contained cases.
- The current dataset has 60 cards that are the same size, shape and case.
- Some of the cards are slightly rotated within the case +-1 degree.
- The deliverable should be C# code that can be added to an existing project.
- The algorithm should work with each item in the dataset.
- The algorithm should take less than 10 seconds but duration is not critical.
- Some of the more difficult items have been attached.
- As a fallback, the algorithm can produce up to 3 choices for a user to manually select from.
- The edges of the contained card should be exact.
- The card should be extracted into a new deskewed, properly oriented image.
- All cards should be very close to the same height and width aspect, but may vary by resolution.
- Some of the cards have very visible edges and are easy to extract with OpenCV.
- Others have backgrounds that are very close to the background and are more difficult to extract.
- Some cards have multiple colors, making the borders harder to detect.
- Addtional projects will be created after this one is complete. It is part of a processing workflow.