Detecting skin in images by using different color spaces.
This is an implementation of the paper down below:
https://www.atlantis-press.com/proceedings/iccasp-16/25871632
Here is the main part of the article in which the algorithm is explained.
The algorithm described in the paper "Human Skin Detection Using RGB, HSV and YCbCr Color Models" involves a multi-step process for skin detection in images. The algorithm includes the following key steps:
- Conversion of ARGB values to HSV and YCbCr values using conversion factors and built-in functions.
- Comparison of the HSV, YCbCr, and ARGB values of each pixel with standard values of a skin pixel.
- Decision-making based on whether the values lie within predefined threshold ranges for each parameter in the different color spaces (HSV, YCbCr, and ARGB).
As there can be seen in the algorithm, there is an OR operation, by which the algorithm divides into two steps. So I decided to make two masks, one for the conditions stated before OR, and one for the conditions stated after OR; and at the end, the union of them is computed and this final mask is applied on the original image.
Additionaly, morthological operations and other noise removal methods have been applied in order to improve the quality of the masks.
sample result:
