Hi all,

I obtained some pictures using camera. It looks to me there are some 
additive noise that was added to all image pixel components: from what I've 
observed, there are many small fluctuations/peaks/spikes of height 10-30, 
while the useful image region has values about 300.

I am afraid that such noise has also been added to the useful image region, 
so the images have been universally shifted by 10-30 noise values. How can I 
estimate the noise and substract the noise value from those useful image 
region? Can I take the average of 10 and 30 and get 20, and then substract 
the 20 from all images?

Any better ideas?

thanks a lot 

kiki wrote:

> Hi all,
> 
> I obtained some pictures using camera. It looks to me there are some 
> additive noise that was added to all image pixel components: from what I've 
> observed, there are many small fluctuations/peaks/spikes of height 10-30, 
> while the useful image region has values about 300.
> 
> I am afraid that such noise has also been added to the useful image region, 
> so the images have been universally shifted by 10-30 noise values. How can I 
> estimate the noise and substract the noise value from those useful image 
> region? Can I take the average of 10 and 30 and get 20, and then substract 
> the 20 from all images?
> 
> Any better ideas?
> 
> thanks a lot 

Are the individual pixel shifts nearly the same in each of several images?

Jerry
-- 
Engineering is the art of making what you want from things you can get.
¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯

"Jerry Avins" <j...@ieee.org> wrote in message 
news:2...@uni-berlin.de...
> kiki wrote:
>
>> Hi all,
>>
>> I obtained some pictures using camera. It looks to me there are some
>> additive noise that was added to all image pixel components: from what 
>> I've
>> observed, there are many small fluctuations/peaks/spikes of height 10-30,
>> while the useful image region has values about 300.
>>
>> I am afraid that such noise has also been added to the useful image 
>> region,
>> so the images have been universally shifted by 10-30 noise values. How 
>> can I
>> estimate the noise and substract the noise value from those useful image
>> region? Can I take the average of 10 and 30 and get 20, and then 
>> substract
>> the 20 from all images?
>>
>> Any better ideas?
>>
>> thanks a lot
>
> Are the individual pixel shifts nearly the same in each of several images?
>
> Jerry
> -- 
> Engineering is the art of making what you want from things you can get.
> ¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯

Hi Jerry,

The problem is that I don't know what should those shifts be and what should 
the correct pixel values be...

I just see there are small fluctuation/spikes/peaks everywhere, but are 
locally rapid fluctuation.

How do I do that?

Thanks a lot! 

kiki wrote:

> "Jerry Avins" <j...@ieee.org> wrote in message 

  ...

> Hi Jerry,
> 
> The problem is that I don't know what should those shifts be and what should 
> the correct pixel values be...
> 
> I just see there are small fluctuation/spikes/peaks everywhere, but are 
> locally rapid fluctuation.
> 
> How do I do that?

Take two pictures of a blank wall without moving the camera. Subtract
one from the other. Don't rely on the camera not moving at all if the
result isn't mostly zero, but shift one and two pixels in each direction
(15 total shifted positions in all). If any of those are mostly zero,
you have a correction mask. Otherwise, look at the spectrum of the
noise. If there is a low-pass filter that can block it without degrading
the picture too much, good. If not, that's hard luck. To filter, you
needn't necessarily work with the FT of the image. The chances are that
a 3 by 3 kernel (or at most a 5 by 5 with the corners missing) will do.

Jerry
-- 
Engineering is the art of making what you want from things you can get.
¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯

If the noise isn't simply additive white gaussian noise, but actually 
exhibits some systematic behavior, there might be a way out if you can 
find a way to model the statistics of your noise. Now, that's gonna 
require some hardcore signal processing, and I'm not sure that's your 
game of choice....

/Johan

Jerry Avins wrote:

> kiki wrote:
> 
> 
>>"Jerry Avins" <j...@ieee.org> wrote in message 
> 
> 
>   ...
> 
> 
>>Hi Jerry,
>>
>>The problem is that I don't know what should those shifts be and what should 
>>the correct pixel values be...
>>
>>I just see there are small fluctuation/spikes/peaks everywhere, but are 
>>locally rapid fluctuation.
>>
>>How do I do that?
> 
> 
> Take two pictures of a blank wall without moving the camera. Subtract
> one from the other. Don't rely on the camera not moving at all if the
> result isn't mostly zero, but shift one and two pixels in each direction
> (15 total shifted positions in all). If any of those are mostly zero,
> you have a correction mask. Otherwise, look at the spectrum of the
> noise. If there is a low-pass filter that can block it without degrading
> the picture too much, good. If not, that's hard luck. To filter, you
> needn't necessarily work with the FT of the image. The chances are that
> a 3 by 3 kernel (or at most a 5 by 5 with the corners missing) will do.
> 
> Jerry

kiki wrote:
> Hi all,
> 
> I obtained some pictures using camera. It looks to me there are some 
> additive noise that was added to all image pixel components: from what I've 
> observed, there are many small fluctuations/peaks/spikes of height 10-30, 
> while the useful image region has values about 300.
> 
> I am afraid that such noise has also been added to the useful image region, 
> so the images have been universally shifted by 10-30 noise values. How can I 
> estimate the noise and substract the noise value from those useful image 
> region? Can I take the average of 10 and 30 and get 20, and then substract 
> the 20 from all images?
> 
> Any better ideas?
> 
> thanks a lot 
> 
> 

Hi,

If you are certain that it is unwanted additive noise, then take all 
values that fall in that range and substitute the 'true' value. I saw 
that  one many years ago on some pictures of saturn and the results were 
stunning.

They called it something like a stairstep function. The output for some 
range of values is a single value.

If you encounter a soft gradient type of image, all bets are off.

Pat

kiki wrote:
> Hi all,
> 
> I obtained some pictures using camera. It looks to me there are some 
> additive noise that was added to all image pixel components: from what I've 
> observed, there are many small fluctuations/peaks/spikes of height 10-30, 
> while the useful image region has values about 300.
> 
> I am afraid that such noise has also been added to the useful image region, 
> so the images have been universally shifted by 10-30 noise values. How can I 
> estimate the noise and substract the noise value from those useful image 
> region? Can I take the average of 10 and 30 and get 20, and then substract 
> the 20 from all images?
> 
> Any better ideas?

Image acquisition is everything.  Take several images
of the same scene and then average them together.
The random noise will be different in each image, and
so will average away.

This is called ensemble averaging, and there is no
better way to reduce random noise in an image.

-- 
Regards,
Martin Leese
E-mail: p...@see.Web.for.e-mail.INVALID
Web: http://members.tripod.com/martin_leese/

Hi,

If your main goal is just to improve these images you might take a look at a very nice tool that has a demo: Neat Image
at http://www.neatimage.com
It has saved the day on some of my photographs.

Tom

In article <cmofj5$5dn$1...@news.Stanford.EDU>, "kiki" <l...@yahoo.com> writes:
>
>I obtained some pictures using camera. It looks to me there are some 
>additive noise that was added to all image pixel components: from what I've 
>observed, there are many small fluctuations/peaks/spikes of height 10-30, 
>while the useful image region has values about 300.
>
>I am afraid that such noise has also been added to the useful image region, 
>so the images have been universally shifted by 10-30 noise values. How can I 
>estimate the noise and substract the noise value from those useful image 
>region? Can I take the average of 10 and 30 and get 20, and then substract 
>the 20 from all images?
>
>Any better ideas?
>


Tom
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
     Tom Kinter   t...@mayo.edu   507-284-4981
      Mayo Foundation Rochester MN 55905 USA
          http://www.mayo.edu/ultrasound
     Tom's Place http://www.tkinter.smig.net
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~