2
$\begingroup$

I have an several images that look like the following:

enter image description here

I would like to separate the skin from the trousers here without specifically training a segmentation model. Can anyone suggest any such algorithm that can potentially achieve this task? Also a dummy example that can demonstrate this would be great. Thank you

Improve this question
$\endgroup$
2

1 Answer 1

Reset to default
0
$\begingroup$

Given there is no learning model which can combine spatial and color information you're left with thresholding based operations.

The tricks here are limited to the color space and the calibration tables of different methods (See Comparative Study of Skin Color Detection and Segmentation in HSV and YCbCr Color Space which is available on SciHub).

Now you can combine several approaches either with OR to increase the recall or AND to increase the precision.

I'd choose the AND path and then add another step for segmentation which is similar to A Closed Form Solution to Natural Image Matting to fine tune the skin mask.

Improve this answer
$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.