EM for gamma mixture model for image histogram

I have implemented the (Expectation Maximization )EM for gamma mixtures to model an image histogram . When I plotted the PDF using the estimated parameters  , it is differing from the original image histogram. Some URLs pointed that the intialization step of the parameters playing a key role in EM. My question is how I intialize these parameters?
00matrix00Asked:
Who is Participating?

[Webinar] Streamline your web hosting managementRegister Today

x
 
TommySzalapskiConnect With a Mentor Commented:
Note that you are creating an estimate so your values should not be expected to match perfectly. They should be close though. Depending on what you start with as an initial guess it will take more iterations to get to the solution, but you should get to the same solution no matter what.
Don't forget that the process is iterative. So you do the expectation step and maximization step over and over until the values converge (you stop getting new ones). So your values should do something like this:
.5 <- initial guess
1.72222 <-after one run
1.53232
1.63451
1.62531
1.62492
When the new value is close enough (for you) to the previous value, you stop and use that.
0
 
mlmccCommented:
This question has been classified as abandoned and is closed as part of the Cleanup Program. See the recommendation for more details.
0
All Courses

From novice to tech pro — start learning today.