How to normalize an image to zero mean and unit variance Options

Re: How to normalize an image to zero mean and unit variance Options

On Sep 20, 8:58=A0am, lijing  wrote:
> Good everyone,
> I am working on face recognition. I need your help on how to
> normalize
> an image to zero mean and unit variance in MATLAB. In addition, I
> will
> also appreciate simple explanation on it.
>
> Thank you very much.
>
> Lijing

--------------------------------------------
Lijing:
Do a web search for "Wallis Filter" - it's a simple locally adaptive
filter that kind of does edge detection, and does a more general
version of what you say: it lets you specify what mean and standard
deviation you want.  It's also locally adaptive so it forces these
parameters into the window size that you specify, rather than the
entire image (which wouldn't really do any filtering, it would just do
intensity scaling).  Not sure what you're after though, perhaps you
can give your motivation, need, or application.  Or post a picture and
tell us what you're trying to find or measure.
Regards,
ImageAnalyst

Re: How to normalize an image to zero mean and unit variance Options

lijing wrote:
> Good everyone,
> I am working on face recognition. I need your help on how to
> normalize
> an image to zero mean and unit variance in MATLAB. In addition, I
> will
> also appreciate simple explanation on it.
> 

Are you using a textbook?

You might find some help in section 4.7 of:

http://www.jgcampbell.com/ip/ip4.txt

I assume that you will know how to sum over all pixels in MATLAB.

Zero mean --- subtract the mean from all pixels.

Unit variance ( = unit standard deviation). Divide all pixels py the 
standard deviation.

As regards negative pixels --- the mathematics may want zero mean and 
hence some negative values; don't worry about the physical 
meaninglessness of that.

A piece of advice. Your pixels may be byte values (0..255), or even 
integer (-(large +1), large). Convert them to floating point for 
calculations.

Best regards,

Jon C.