Help with image thresholding?

I need to get a binary image of the fist (shown below).

I've tried using graythresh to get the threshold value, tried my own threshold values, used imgradient to get the edges and even used edge detection, but I can't seem to get a clear outline of the fist.

I'm using blob analysis for detecting the fist.

Can someone help me out? Really appreciate it!

Answer 1

[Anand]

Try working in the HSV space. Here's what I got

I = imread('http://www.mathworks.com/matlabcentral/answers/uploaded_files/8923/gest2_7.jpg');

% Convert to HSV
I = rgb2hsv(I);

% Threshold based on hue
mask = I(:,:,1)>=0.023 & I(:,:,1)<=0.082;

% Cleanup mask
...

You'll need to play around with the thresholds some more and do some post-processing on the mask like filling holes and smoothing the boundaries.

[Shravanthi]

I'm really sorry, I forgot to mention that the picture is from video from the webcam, whose color space is set to gray scale. Is there any way of doing this without converting to hsv?

[Image Analyst]

Why? Why are you not using color????

[Shravanthi]

Um, because I have to detect other gestures, and blob analysis works fine for all of them, except this one. All the other images are in grayscale too. I just can't find a proper threshold value for this one.

[Image Analyst]

Can you get better illumination of the background, so that you don't have such a severe gradient?

[Shravanthi]

This was the best lighting we could get in the room where we have to show the project. The background is black cardboard. Is there something else I can do?