Extract the hand region from an image

I have an image that looks like:
This image is from the dataset and I implemented skin color extraction to this image.
My goal here is to extract only hand region, that's something like this:
Do you have any idea about how to do so using some image processing algorithms?
For example: to create two rectangle or circle-like shapes that contain same/similar color intensities in neighbouring regions, one for face and the other one for hand and check if those circles contain two little black regions(in this case, probably eyes) or two blackish lines (eyebrows) and extract the second shape.

8 Comments

Will the hand always be curled into a fist?
To detect a hand with in an image, you can use functions from Computer Vision System Toolbox and train using your dataset of hand images.
Click on the following URL to know more about "Object Detection" using Computer Vision System Toolbox:
I trained the dataset via Inception and got terrible results, because the images contain too much unnecessary information like background, person body, etc. Therefore I want to extract only hand region shown above and train only them. Do you think Cascade object detector works well anyway?
It would be a lot easier if you could remove the background clutter. Can you put a black sheet behind all your subjects?
Thank you for response! Which noise removal technique would work best for my case? I don't really understand "putting black sheet" behind subjects! Please, can you clarify things a bit? Thanks
Image Analyst
Image Analyst on 10 Apr 2017
Edited: Image Analyst on 10 Apr 2017
You see that door behind the woman? Well what is that door wasn't there and the only thing behind here was a totally black wall, or a white sheet. Have you ever heard of a cyclorama?
Wouldn't that make it easier to find only the hand? Of course it would.
I see now. I can't do that, because I only have a dataset downloaded from the WEB. In your suggestion, I would have to create my own dataset which is impossible for me. I need to use existing ones and do some processing. Any other suggestions?

Sign in to comment.

Answers (0)

Asked:

on 5 Apr 2017

Commented:

on 10 Apr 2017

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!