Hi all, We used camera to do some imaging and got some raw image data, which are very noisy at the low signal end(where there should be some low signal) or the dark end(where there should be no signal theoratically)... And there are some some spikes on both dark and bright regions. The bright parts of the image are also noisy... Please advise which type of filter can be best for denoising of such images, hopefully the useful image parts will not be affected or minorly affected... Thanks a lot! -L
need advice on image filter?
Started by ●October 31, 2004
Reply by ●October 31, 20042004-10-31
lucy wrote:> Hi all, > > We used camera to do some imaging and got some raw image data, which are > very noisy at the low signal end(where there should be some low signal) or > the dark end(where there should be no signal theoratically)... > > And there are some some spikes on both dark and bright regions. The bright > parts of the image are also noisy... > > Please advise which type of filter can be best for denoising of such images, > hopefully the useful image parts will not be affected or minorly affected... > > Thanks a lot! > > -LBefore you go on, take a few pictures of a featureless object, with the camera held stationary between exposures. Any noise correlation represents defects in the sensor array and van be subtracted out of the images you want to process. Dark noise and highlight noise might be different, but taking advantage of that might be hard. Jerry -- Engineering is the art of making what you want from things you can get. �����������������������������������������������������������������������
Reply by ●November 1, 20042004-11-01
"Jerry Avins" <jya@ieee.org> wrote in message news:2ulnp6F2d39jaU1@uni-berlin.de...> lucy wrote: >> Hi all, >> >> We used camera to do some imaging and got some raw image data, which are >> very noisy at the low signal end(where there should be some low signal) >> or >> the dark end(where there should be no signal theoratically)... >> >> And there are some some spikes on both dark and bright regions. The >> bright >> parts of the image are also noisy... >> >> Please advise which type of filter can be best for denoising of such >> images, >> hopefully the useful image parts will not be affected or minorly >> affected... >> >> Thanks a lot! >> >> -L > > Before you go on, take a few pictures of a featureless object, with the > camera held stationary between exposures. Any noise correlation > represents defects in the sensor array and van be subtracted out of the > images you want to process. Dark noise and highlight noise might be > different, but taking advantage of that might be hard. > > Jerry > -- > Engineering is the art of making what you want from things you can get. > �����������������������������������������������������������������������Hi Jerry, That's a very good point. We have not thought about that... yeah, we should take a picture of completely dark NULL object and then do a subtraction: picture with meaningful object images - picture with NULL object = noise-reduced image... Then what do we do after that? Use what filter to reduce noise? Thanks a lot! L
Reply by ●November 1, 20042004-11-01
Jerry Avins <jya@ieee.org> wrote in message news:<2ulnp6F2d39jaU1@uni-berlin.de>...> lucy wrote: > > Hi all, > > > > We used camera to do some imaging and got some raw image data, which are > > very noisy at the low signal end(where there should be some low signal) or > > the dark end(where there should be no signal theoratically)... > > > > And there are some some spikes on both dark and bright regions. The bright > > parts of the image are also noisy... > > > > Please advise which type of filter can be best for denoising of such images, > > hopefully the useful image parts will not be affected or minorly affected... > > > > Thanks a lot! > > > > -L > > Before you go on, take a few pictures of a featureless object, with the > camera held stationary between exposures.If the camera has a cap for the objective cap, take some pictures with the lense cap on, with the same exposure times as you used for the "real" pictures. That way you could get an impression of the eigen noise of the CCD, and you might be able to subtract it from your images.> Any noise correlation > represents defects in the sensor array and van be subtracted out of the > images you want to process. Dark noise and highlight noise might be > different, but taking advantage of that might be hard. > > JerryRune
Reply by ●November 1, 20042004-11-01
On 2004-11-01 10:07:12 +0100, allnor@tele.ntnu.no (Rune Allnor) said:> If the camera has a cap for the objective cap, take some pictures > with the lense cap on, with the same exposure times as you used for the > "real" pictures. That way you could get an impression of the eigen > noise of the CCD, and you might be able to subtract it from your images. > > RuneOr for that matter, if you can do this simply take several pictures of the same scene and average them together. Works for static scenes only, obviously, but is a fairly good method to minimize noise. -- Stephan M. Bernsee http://www.dspdimension.com
Reply by ●November 1, 20042004-11-01
lucy wrote: ...> Then what do we do after that? Use what filter to reduce noise?Sorry, I only know the easy stuff. In any case, the "best" filter is determined by the natures of the noise and image. and by processing resources. Good luck. Jerry -- Engineering is the art of making what you want from things you can get. �����������������������������������������������������������������������
Reply by ●November 1, 20042004-11-01
Stephan M. Bernsee wrote:> On 2004-11-01 10:07:12 +0100, allnor@tele.ntnu.no (Rune Allnor) said: > >> If the camera has a cap for the objective cap, take some pictures >> with the lense cap on, with the same exposure times as you used for >> the "real" pictures. That way you could get an impression of the eigen >> noise of the CCD, and you might be able to subtract it from your images. >> >> Rune > > > Or for that matter, if you can do this simply take several pictures of > the same scene and average them together. Works for static scenes only, > obviously, but is a fairly good method to minimize noise.That works if the noise is random and the scene is static. We were discussing removal of the relatively simple static noise as a pre-processing step. Jerry -- Engineering is the art of making what you want from things you can get. �����������������������������������������������������������������������
Reply by ●November 1, 20042004-11-01
"Stephan M. Bernsee" <spam@dspdimension.com> wrote in message news:2umcuqF2b6ekoU1@uni-berlin.de...> On 2004-11-01 10:07:12 +0100, allnor@tele.ntnu.no (Rune Allnor) said: > > > If the camera has a cap for the objective cap, take some pictures > > with the lense cap on, with the same exposure times as you used for the > > "real" pictures. That way you could get an impression of the eigen > > noise of the CCD, and you might be able to subtract it from your images. > > > > Rune > > Or for that matter, if you can do this simply take several pictures of > the same scene and average them together. Works for static scenes only, > obviously, but is a fairly good method to minimize noise. > -- > Stephan M. Bernsee > http://www.dspdimension.com >Hello Stephan and others, The averaging together of multiple images is good for reducing thermal noise, but CCDs sometimes develop "stuck pixels" with long exposures. So a common trick is to do a long dark exposure and locate the stuck pixels via thresholding and then black the corresponding ones out in the actual image. This is how the Nikon D100's noise reduction for long exposures works. Clay
Reply by ●November 1, 20042004-11-01
On 2004-11-01 15:07:16 +0100, Jerry Avins <jya@ieee.org> said:> That works if the noise is random and the scene is static. We were > discussing removal of the relatively simple static noise as a > pre-processing step. > > JerryThe most common noise type would be thermal noise, for which this method works well. Since we don't know what the OP is trying to accomplish, I don't know if he has to deal with moving objects and whether his noise is thermal or not... maybe he can tell us. -- Stephan M. Bernsee http://www.dspdimension.com
Reply by ●November 1, 20042004-11-01
That's what my digital camera (and many others) do automatically. On the Canon G5, for any exposure longer than 1 second, it takes the picture normally, and then takes another identical exposure with the shutter closed and then does some filtering to remove the noise. I presume it could be as simple as just subtracting the second from the first, but it might be something more complex. "Jerry Avins" <jya@ieee.org> wrote in message news:2ulnp6F2d39jaU1@uni-berlin.de...> lucy wrote: > > Hi all, > > > > We used camera to do some imaging and got some raw image data, which are > > very noisy at the low signal end(where there should be some low signal) or > > the dark end(where there should be no signal theoratically)... > > > > And there are some some spikes on both dark and bright regions. The bright > > parts of the image are also noisy... > > > > Please advise which type of filter can be best for denoising of such images, > > hopefully the useful image parts will not be affected or minorly affected... > > > > Thanks a lot! > > > > -L > > Before you go on, take a few pictures of a featureless object, with the > camera held stationary between exposures. Any noise correlation > represents defects in the sensor array and van be subtracted out of the > images you want to process. Dark noise and highlight noise might be > different, but taking advantage of that might be hard. > > Jerry > -- > Engineering is the art of making what you want from things you can get. > �����������������������������������������������������������������������
