bwconncomp and conn=3-by-3 constant marix not working for C/C++ Code Generation, despite using R2023b

Hello,

I want to use bwconncomp with 2D binary images and pure column-wise connectivity for code generation, and was delighted to read in the documentation, that since MATLAB R2023a the limitation for conn=4 and conn=8 is supposed to be over:

R2023a: Additional support for C code generation

You can now generate C code for input images of any dimension. Before, bwconncomp supported code generation only for 2-D inputs. Also, you can now specify any connectivity that is a compile-time constant. Before, bwconncomp supported code generation only for connectivities of 4, 8, [0 1 0;1 1 1;0 1 0], or ones(3)

However, this line just does not want to pass in the coder

CC = bwconncomp(BW,[0 1 0;0 1 0;0 1 0]);

and always produces an error "bwconncomp>3, Line 25: Expected a scalar. Non-scalars are not supported with logical operators...."

despite using MATLAB version R2023b, where as

 CC = bwconncomp(BW,[0 1 0;1 1 1;0 1 0]);

just compiles fine (which actually is conn=4). Why is that, if it is supposed to work now for any 3-by-3 connectivities, as alleged in the documentation ?

Thanks

Answer 1

[Githin George]

Hello Robin,

I've attempted to generate C code for the "bwconncomp" function using the "conn" matrix you have provided, but I'm encountering the same error message. Upon examining the "bwconncomp.m" file, I noticed that the error is triggered by a validation check on the "conn" matrix (refer to Line: 25 in bwconncomp.m). The underlying issue is that the "conn" matrix is treated as a compile-time constant, which is essentially a “coder.Constant” object. The validation uses an “ismember” check, which passes for a "double" matrix but fails for compile-time constants. In the scenario where conn = [0 1 0; 1 1 1; 0 1 0], it gets automatically converted to “conn = 4”. This conversion allows code generation to complete successfully.

For further information on the “coder.Constant” class and "ismember" function, please refer to the documentation links below:

https://www.mathworks.com/help/fixedpoint/ref/coder.constant-class.html

https://www.mathworks.com/help/matlab/ref/double.ismember.html

Given that the error originates from a toolbox file, I recommend that you contact the MathWorks Technical Support team for additional support. You can reach out to them using the following link:

https://www.mathworks.com/support/contact_us.html

I hope this information is helpful to you.