Gaussian Distribution for image pixels

Hi all

Can any one help me how to model the variation of each pixel in grayscale image using its mean and std

as there is one function in matlab normpdf.

y=normpdf(img,mean(img(:),std(img(:))));

Any help is appreciated..

Thanks.

[Image Analyst]

I have no idea what you want to do. Maybe you have a sequence of images and you want the mean and standard deviation of each pixel, so you'd have two images - the mean image (over all the input images), and the standard deviation image. But I really have no idea what you want.

[Algorithms Analyst]

I am modeling and subtracting the background from video sequence using 2-dimensional principle component anaylsis. I have extract the foreground pixels as

        Fg(i)= |b1(i)-b2(i)|
Where b1 is original image and b2 is reconstructed background image using PCA. Now this Fg(i) should be Fg>th(i).

Now threshold can be computed from the

th(i) = max (thMin,min (thMax, m(i) +κσ(i))),

where thMax=60 thMin=30 and k=5.

and m is mean and sigma is variance which are obtained from distance between original and reconstructed image using Gaussian distribution.

This threshold is also updating by updating the mean and variance(standard deviation).like that

mnew(i) = (1 −α)m(i) +α||bi−bˆi||

and

σ2new(i)= (1 −α)σ2 (i)+α(1−α)(|bi−bˆi| − m(i))2.

So my problem is computing the initial mean and variance which are obtained from Fg image.

Thanks

[Image Analyst]

What is the difficulty in adding up the 5 previous frames? You have to convert to single or double of course, but that's really straightforward.

[Algorithms Analyst]

how can I update the threshold equations.

because if I compute the mean it is just row vector i want to compute the mean of each pixel in bgframe.

[Algorithms Analyst]

As I have declared the first frame for background modelling and rest of other are the current input frames for object detection.