Create moving average filter

Hi, I need some help in writing a code for the moving average filter but without using any of the existing matlab functions. It's going to be applied to black and white images, 256x256 pixels, with N rows and M columns:

 f(i,j),   i=1,2,…,N and j=1,2,…,M   f(i,j)=[0,255]

If the function is applied to each of the rows of the image, represented by the array:

 [f1 f2 f3 … fk-1 fk fk+1 … fM]

And with the corresponding exit array:

 [g1 g2 g3 … gk-1 gk gk+1 … gM]

How to solve this, in order to create the moving average filter?:

 gk=(agk-1+bgk+cgk+1)/(a+b+c)    for k=2,3,…,M-1 and a=b=c=1

(and to make sure the length of the arrays is the same, g1= f1 e gM = fM )

Thankyou

[bym]

"without using any of the existing matlab functions" is vague

[LM]

Sorry, I mean the filter function.

Answer 1

[Image Analyst]

Sounds like a class assignment. Think about it. Do you think you could do it with 4 for loops? Maybe one pair that visits every pixel in the image (iterates over rows and columns), and another pair inside that visits every pixel in your NxM window that is centered over the pixel you're visiting?

[LM]

Yes, I wasn't sure on the structure I should use here, but thanks for your suggestion, I'll try to do it :)

Answer 2

[Jan]

The moving average can be calculated for the columns at first and the rows afterwards.

For the moving average along the rows or columns a single loop is sufficient: One vector is added to the accumulator and one vector is subtracted.

[Walter Roberson]

Though that does have the disadvantage of having cumulative round-off error.

[Jan]

@Walter: That's correct, but for a 256x256 pixel image this will very likely not matter. If it is a 8 bit grey scale or 8x3 bit RGB image: Using an accumulator of type DOUBLE will be sufficient - as long as the picture are small enough.