I'm trying to implement an algorithm that show how to detect wrinkles in clothes, without going to much into details this is the step that is confusing me.

Assuming we have an image $I$ and a set of angles $\alpha \in \left\{j45^o : 0 \leq j \leq 7 \right\}$ a set of Gabor filters are run on $I$, therefore I get images $I_0,\ldots, I_7$. Now I construct an image $J$ whose pixels are defined as

$$J(x,y) = \max_{0\leq j \leq 7} \left\{ I_j(x,y) \right\}$$

You can also store the corrisponding angle. Now.. apparently the next step is to divide the image $J$ in group of ellipses like region, there's no explanation of how to do this, so it must be something very trivial that I'm missing. The question is... given the image above and the "angle map" how can I derive such regions?

I was thinking an approach like clustering pixels by angle and then fitting an ellipse. But maybe there's a simpler approach, or maybe this one would be totally wrong.

Any suggestions?

The implementation has to be done in opencv C++.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.