The Premultiply Convention in Geometric Transformations does not support C/C++ code generation?

Since version R2022b, the Premultiply Convention has been recommended in priority in Geometric Transformations in the IPL(image processing toolbox) and CVT(computer vision toolbox) toolboxes, which is very much in line with industry usage. However, it is strange that some of the recommended new functions do not support C/C++ code generation, while the old ones do. For example, affinetform2d/affinetform3d do not support this, while the old functions affine2d/affine3d support C/C++ code generation. rigidtform2d/rigidtform3d should also support C/C++ generation, as this geometric transformation is very basic and typical. It is hoped that subsequent versions of TMW will enhance this function.

[Kenneth]

That's a really interesting point about the inconsistency between the new and old geometric transformation functions in MATLAB. It's definitely odd that the newer, recommended 'premultiply convention' functions lack C/C++ code generation, especially when the older ones support it. You're right, rigid transformations are fundamental and should absolutely be included. Hopefully, MathWorks addresses this in future releases.

Answer 1

[Qu Cao]

Thank you for reporting this. There is a bug in the documentation. All the geometric transformation objects with the premultiply convention should support C/C++ code generation.

[xingxingcui]

• As a side note, I have found that some apps have not yet migrated to pre-multiply, for example, after aligning two images in registrationEstimator, the automatically exported functions or variables still use the post-multiply form.This is a minor issue, just be aware of it
• The formula in the Algorithms section at the bottom of the help documentation for the function estimateCameraParameters should also be changed to pre-multiply.

[Qu Cao]

Thank you agian for reporting these issues! We will fix them in a future release.

Answer 2

[Benjamin Thompson]

This is just a notational thing right? Why would it affect how the code is generated and executed? The code generation and execution processes should generate the most efficient code possible.