Re: [SIP-devel] function pifilter

[DRUEL J. <joc...@li...>]

>
> Hello,
>
> I have included your pifilter function into SIP-dev, it is working
> fine.
>
> It is interesting that I tested the function on natural luminance images
> messed up with imnoise, option "salt & pepper".
> The result is quite good, because the edges are preserved, more
> so than median filtering, although some noises are not fully eliminated.
>
> The idea of treating a normal intensity image as a phase image seems
> to makes sense in this case. The pi filter will do an adaptive blur
> of the image, with the blur strength somewhat inversely proportional
> to the variance of the data around the mean. The most abrupt
> variation, the weaker the blur. A variation from maximum to minimum
> distributed along the neighbourhood of a pixel will not blur the
> pixel.
>
> If you think this makes sense, the function pifilter will be useful not
> only for photonics images, but for many other noisy images.

Hi,
All right, I think it makes sense. 
In fact, the first time (and only time) I met this filter was in 
documentation about phased images. But it could be used to remove noise in 
other pictures. That's a good idea.

I let you complete the manual page with the explanations above. Is that OK ?


In the function, I hesitated using the function "normal":
I could have written 
a=(inm*2*%pi/255)-%pi;
instead of  a=normal(inm,255,0);

But this causes a problem if max(inm)>255.

For phased images, the use of the "normal" function is not a problem as these 
images occupy the full range 0-255.
For other images, I don't know if it causes a problem to anybody: the final 
image will be 0-255, whatever the initial image is.
For me, it's not a problem as we can use again the "normal" function to 
adjust the desired range.
So, for me, it's allright to use the normal function, but I don't for the 
other users.

Well, maybe I'm simply asking myself too many questions....

Best regards
Jocelyn


-- 
Powered by Linux