On Jun 25, 2:36=A0pm, "Nasser Abbasi" <n...@[EMAIL PROTECTED]
> wrote:
> hello;
>
> I have 2 images (of same size), both gray levels. =A0Say A and B. =A0B
is=
a
> modified version of A, and I'd like to measure how far B is from A.
>
> I currently use RMSE and also use relative error,absolute error and
> histogram difference. But I 'like' RMSE the most so far.
>
> Is there a better method to quantify, in one number, the difference
betwe=
en
> 2 images that I am overlooking ?
>
> thanks,
> Nasser
It is often desirable to eliminate differences in overall brightness
and/or measure of centrality from the images which typically arise
from lighting and/or exposure changes.
If you designate A as the template image,
min || A - a*B ||
or
min || A - (a*B + b) ||
where a and b are scalars representing brightness scale and centrality
****ft, respectively, that minimize the expressions and ||.||
designates some norm, (not necessarily the L2 norm) provides such
difference measures.
|
