32-Bit Tiff Single Pixel Noise Removal

[Francis]

Hi,

I need to remove single pixel noise from 1000's of 32-bit TIFF images. I tried to use the outlier process but its not working as I have to manually set the threshold. I think I need to automatically apply a known 95 or 99% threshold to the minimum and maximum of the histogram data before I can properly visualize the images.

I am attaching an example of the type of image and its histogram. I hope someone on the list could advise on a batch type macro that could solve this problem.

Best,

Francis


--
ImageJ mailing list: http://imagej.nih.gov/ij/list.html

[Herbie-4]

Good day Francis,

before you start thinking about noise removal, try to define what is
signal and what is noise. If this can't be done in a formal/mathematical
way, you won't be able to successfully and automatically process your data.

Best

Herbie

::::::::::::::::::::::::::::::::::::::::
On 03.02.15 03:26, Francis OBrien wrote:

> Hi,
>
> I need to remove single pixel noise from 1000's of 32-bit TIFF
> images. I tried to use the outlier process but its not working as I
> have to manually set the threshold. I think I need to automatically
> apply a known 95 or 99% threshold to the minimum and maximum of the
> histogram data before I can properly visualize the images.
>
> I am attaching an example of the type of image and its histogram. I
> hope someone on the list could advise on a batch type macro that
> could solve this problem.
>
> Best,
>
> Francis
>
>
> -- ImageJ mailing list: http://imagej.nih.gov/ij/list.html
>

--
ImageJ mailing list: http://imagej.nih.gov/ij/list.html

[Michael Schmid]

Hi Francis,

it depends what you want to do with the images. If the pixel intensities are not important (e.g. you only want to detect the position of the elliptical areas) the easiest would be using the logarithm of the image (Process>Math>Log).

I don't think that the excessively high pixel values are noise. Essentially all of the structure in the image seems to be made of streaky features similar to the very bright 'noise'.

Michael
________________________________________________________________
On Feb 3, 2015, at 03:26, Francis OBrien wrote:

> Hi,
>
> I need to remove single pixel noise from 1000's of 32-bit TIFF images. I tried to use the outlier process but its not working as I have to manually set the threshold. I think I need to automatically apply a known 95 or 99% threshold to the minimum and maximum of the histogram data before I can properly visualize the images.
>
> I am attaching an example of the type of image and its histogram. I hope someone on the list could advise on a batch type macro that could solve this problem.
>
> Best,
>
> Francis

--
ImageJ mailing list: http://imagej.nih.gov/ij/list.html

[Francis]

Herbie,

I would like to threshold the images based on a 95% or 99% confidence interval. The single pixel that I want to remove is a detector counting artifact of a mass spectrometer. Any suggestions?

Best,

Francis

--
ImageJ mailing list: http://imagej.nih.gov/ij/list.html

[Francis]

Michael,

The single hot pixels are orders of magnitude than the rest of the image and it is an artifact based on a the counting circuit of a mass spectrometer. The streakiness of the this raw image is due to the transport of aerosol into the mass spectrometer.

How can I threshold out this hot pixel artifact so we can start analyzing the data.

Best,

Francis



--
ImageJ mailing list: http://imagej.nih.gov/ij/list.html

[Saalfeld, Stephan]

Hi Francis,

if the hotpixel intensities are far enough from the local intensity
distribution, you may be successful with the Fiji plugin

Plugins>Integral Image Filters>Remove Outliers

for every pixel, the plugin estimates the mean and standard deviation in
a rectangle (block) of given size, and if the pixels falls beyond mean +
alpha * std, then it replaces it by the average of its direct neighbors,
i.e. smears it away.  alpha and block-size are parameters of the plugin,
it has a preview button and is very fast, independent of the block size.

Best,
Stephan



On Wed, 2015-02-04 at 22:28 -0500, Francis OBrien wrote:

> Michael,
>
> The single hot pixels are orders of magnitude than the rest of the image and it is an artifact based on a the counting circuit of a mass spectrometer. The streakiness of the this raw image is due to the transport of aerosol into the mass spectrometer.
>
> How can I threshold out this hot pixel artifact so we can start analyzing the data.
>
> Best,
>
> Francis
>
>  
>
> --
> ImageJ mailing list: http://imagej.nih.gov/ij/list.html

--
ImageJ mailing list: http://imagej.nih.gov/ij/list.html

[Francis]

Michael,

Thank you for your suggestion. When I look at the data in log view it
is exactly what I would expect. But the pixels that are distorting my view
need to be highlighted and removed as instrument artifacts. Any ideas?

Best,

Francis

On Tue, Feb 3, 2015 at 9:16 AM, Michael Schmid <[hidden email]>
wrote:

> Hi Francis,
>
> it depends what you want to do with the images. If the pixel intensities
> are not important (e.g. you only want to detect the position of the
> elliptical areas) the easiest would be using the logarithm of the image
> (Process>Math>Log).
>
> I don't think that the excessively high pixel values are noise.
> Essentially all of the structure in the image seems to be made of streaky
> features similar to the very bright 'noise'.
>
> Michael
> ________________________________________________________________
> On Feb 3, 2015, at 03:26, Francis OBrien wrote:
>
> > Hi,
> >
> > I need to remove single pixel noise from 1000's of 32-bit TIFF images. I
> tried to use the outlier process but its not working as I have to manually
> set the threshold. I think I need to automatically apply a known 95 or 99%
> threshold to the minimum and maximum of the histogram data before I can
> properly visualize the images.
> >
> > I am attaching an example of the type of image and its histogram. I hope
> someone on the list could advise on a batch type macro that could solve
> this problem.
> >
> > Best,
> >
> > Francis
>
> --
> ImageJ mailing list: http://imagej.nih.gov/ij/list.html
>

--
Regards,

Francis

Research graduate,
New York School of Medicine
New York

--
ImageJ mailing list: http://imagej.nih.gov/ij/list.html

[Michael Schmid]

Hi Francis,

it would be the best if you mark a few pixels that you don't like in the log image. Otherwise, I have no idea what you consider bad pixels.

Michael
________________________________________________________________
On Feb 5, 2015, at 19:42, Francis OBrien wrote:

> Michael,
>
> Thank you for your suggestion. When I look at the data in log view it
> is exactly what I would expect. But the pixels that are distorting my view
> need to be highlighted and removed as instrument artifacts. Any ideas?
>
> Best,
>
> Francis
>
> On Tue, Feb 3, 2015 at 9:16 AM, Michael Schmid <[hidden email]>
> wrote:
>
>> Hi Francis,
>>
>> it depends what you want to do with the images. If the pixel intensities
>> are not important (e.g. you only want to detect the position of the
>> elliptical areas) the easiest would be using the logarithm of the image
>> (Process>Math>Log).
>>
>> I don't think that the excessively high pixel values are noise.
>> Essentially all of the structure in the image seems to be made of streaky
>> features similar to the very bright 'noise'.
>>
>> Michael
>> ________________________________________________________________
>> On Feb 3, 2015, at 03:26, Francis OBrien wrote:
>>
>>> Hi,
>>>
>>> I need to remove single pixel noise from 1000's of 32-bit TIFF images. I
>> tried to use the outlier process but its not working as I have to manually
>> set the threshold. I think I need to automatically apply a known 95 or 99%
>> threshold to the minimum and maximum of the histogram data before I can
>> properly visualize the images.
>>>
>>> I am attaching an example of the type of image and its histogram. I hope
>> someone on the list could advise on a batch type macro that could solve
>> this problem.
>>>
>>> Best,
>>>
>>> Francis
>>
>> --
>> ImageJ mailing list: http://imagej.nih.gov/ij/list.html
>>
>
>
>
> --
> Regards,
>
> Francis
>
> Research graduate,
> New York School of Medicine
> New York
>
> --
> ImageJ mailing list: http://imagej.nih.gov/ij/list.html

--
ImageJ mailing list: http://imagej.nih.gov/ij/list.html