# image correlation and autocorrelation

Started by February 17, 2009
```Hello forum,

this is a simple one:

I found two forms of autocorrelation for 1D signals:
biased and unbiased.
Does anyone know the difference?

Also, all the signals or images we work with are finite in length.
If autocorrelation is carried out, we need to shift the image around with
respect to itself.......( that would correspond to the lag (shift) in
space).
In order to do that, do we first need to pad the images with a lot of
zeros?

If my image is NxN, I guess the zero padded image should be 3Nx3N, to make
space for the lateral shift of the identical copy, right?

thanks
fisico30
```
```On Feb 17, 9:55&#2013266080;am, "fisico30" <marcoscipio...@gmail.com> wrote:
> Hello forum,
>
> this is a simple one:
>
> I found two forms of autocorrelation for 1D signals:
> biased and unbiased.
> Does anyone know the difference?
>
> Also, all the signals or images we work with are finite in length.
> If autocorrelation is carried out, we need to shift the image around with
> respect to itself.......( that would correspond to the lag (shift) in
> space).
> In order to do that, do we first need to pad the images with a lot of
> zeros?
>
> If my image is NxN, I guess the zero padded image should be 3Nx3N, to make
> space for the lateral shift of the identical copy, right?
>
> thanks
> fisico30

http://lmgtfy.com/?q=autocorrelation+wikipedia&l=1
```
```On 17 Feb, 15:55, "fisico30" <marcoscipio...@gmail.com> wrote:
> Hello forum,
>
> this is a simple one:
>
> I found two forms of autocorrelation for 1D signals:
> biased and unbiased.

No, you didn't. What you found was two different
*estimators* for the autocorrelation.

Rune
```