ImageJ

Clearing 32 bit float images

_‹ Previous Topic Next Topic _›

Classic

List

Threaded

6 messages Options

Burger Wilhelm

Clearing 32 bit float images

Hello all,

when applying 'Edit -> Clear' on a 32 bit float image (FloatProcessor) I would assume that the affected pixel values are set to zero. However, the command inserts some other (display-related?) value. E.g., on the "bridge" sample image, the new value is 85.0.

Is this intended behaviour? If yes, how is the clear value determined?

--Wilhelm

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

Gabriel Landini

Re: Clearing 32 bit float images

On Tuesday 02 Jun 2015 10:09:55 Burger Wilhelm wrote:
> when applying 'Edit -> Clear' on a 32 bit float image (FloatProcessor) I
> would assume that the affected pixel values are set to zero. However, the
> command inserts some other (display-related?) value. E.g., on the "bridge"
> sample image, the new value is 85.0.
>
> Is this intended behaviour? If yes, how is the clear value determined?

Hi Wilhelm,
It uses the values in the colour picker.
If you set the background colour to 0 it clears with that value.

Cheers

Gabriel

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

Burger Wilhelm

Re: Clearing 32 bit float images

Hello Gabriel,

yes, this works, thanks!

It seems odd though - at least for scalar-valued images - that the 'clear' value must be specified as an RGB tuple. And is there any way to set the background to a negative value?

The meaning of "Invert" on float images seems to be a similar issue ...

--Wilhelm

> -----Original Message-----
> From: ImageJ Interest Group [mailto:[hidden email]] On Behalf Of
> Gabriel Landini
> Sent: Tuesday, June 02, 2015 10:54 AM
> To: [hidden email]
> Subject: Re: Clearing 32 bit float images
>
> On Tuesday 02 Jun 2015 10:09:55 Burger Wilhelm wrote:
> > when applying 'Edit -> Clear' on a 32 bit float image (FloatProcessor) I
> > would assume that the affected pixel values are set to zero. However, the
> > command inserts some other (display-related?) value. E.g., on the "bridge"
> > sample image, the new value is 85.0.
> >
> > Is this intended behaviour? If yes, how is the clear value determined?
>
> Hi Wilhelm,
> It uses the values in the colour picker.
> If you set the background colour to 0 it clears with that value.
>
> Cheers
>
> Gabriel
>
> --
> ImageJ mailing list: http://imagej.nih.gov/ij/list.html

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

Michael Schmid

Re: Clearing 32 bit float images

Hi Wilhelm,

for setting a pixel value to zero (or any other [uncalibrated] numeric value): Use Process>Math>Set.
If you have a float image and you want to invert it by changing the sign, use Process>Math>Multiply to multiply by -1.

The Edit>Fill (Edit>Clear) command uses the eyedropper foreground (background) color or whatever comes closest for the way the image is currently displayed.

Edit>Invert uses the current display range, i.e. the min and max that you can set in the Brightness&Contrast panel. So you can have e.g. a 16-bit image, display it as 12-bit image (0-4095) and have 'Invert' create the correct values if if you would invert a 12-bit image (0 becomes 4095, and 4095 becomes 0). You could also have a float image with pixel values 0-255, display it with this range, and do the 'Invert' operation just as if it were an 8-bit image, but with floating-point accuracy.

With this convention, Edit>Invert also ensures that an image that is displayable with the current B&C settings remains displayable after 'Invert'. If you have a floating-point image with pixels e.g. 0-255, and you would invert it by multiplying with -1, it will be all black when shown with the 0-255 display range.

Michael
________________________________________________________________
On Jun 2, 2015, at 11:26, Burger Wilhelm wrote:

> Hello Gabriel,
>
> yes, this works, thanks!
>
> It seems odd though - at least for scalar-valued images - that the 'clear' value must be specified as an RGB tuple. And is there any way to set the background to a negative value?
>
> The meaning of "Invert" on float images seems to be a similar issue ...
>
> --Wilhelm
>
>
>> -----Original Message-----
>> From: ImageJ Interest Group [mailto:[hidden email]] On Behalf Of
>> Gabriel Landini
>> Sent: Tuesday, June 02, 2015 10:54 AM
>> To: [hidden email]
>> Subject: Re: Clearing 32 bit float images
>>
>> On Tuesday 02 Jun 2015 10:09:55 Burger Wilhelm wrote:
>>> when applying 'Edit -> Clear' on a 32 bit float image (FloatProcessor) I
>>> would assume that the affected pixel values are set to zero. However, the
>>> command inserts some other (display-related?) value. E.g., on the "bridge"
>>> sample image, the new value is 85.0.
>>>
>>> Is this intended behaviour? If yes, how is the clear value determined?
>>
>> Hi Wilhelm,
>> It uses the values in the colour picker.
>> If you set the background colour to 0 it clears with that value.
>>
>> Cheers
>>
>> Gabriel
>>
>> --
>> ImageJ mailing list: http://imagej.nih.gov/ij/list.html
>
> --
> ImageJ mailing list: http://imagej.nih.gov/ij/list.html

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

Rasband, Wayne (NIH/NIMH) [E]

Re: Clearing 32 bit float images

In reply to this post by Burger Wilhelm

On Jun 2, 2015, at 5:26 AM, Burger Wilhelm <[hidden email]> wrote:
>
> Hello Gabriel,
>
> yes, this works, thanks!
>
> It seems odd though - at least for scalar-valued images - that the 'clear' value must be specified as an RGB tuple. And is there any way to set the background to a negative value?

I do not know how to set the background to a negative value using the GUI but you can use the ImageProcessor.setColor(double) method to set it, as in the following example.

> The meaning of "Invert" on float images seems to be a similar issue …

Each pixel in the float image is inverted using

p = max - (p - min);

where ‘min’ and ‘max’ are the limits of the display range, set using ImageProcessor.setMinAndMax(double,double), as in the example below.

-wayne

imp = IJ.createImage("original", "32-bit ramp", 500, 500, 1);
roi = new OvalRoi(106,108,259,211);
ip = imp.getProcessor();
ip.setRoi(roi);
ip.setColor(-1);
ip.fill(ip.getMask());
ip.setMinAndMax(-1, 1);
ip.resetRoi();
IJ.log("original: "+ip.getStatistics());
imp.show();
ip2 = ip.duplicate();
ip2.invert();
IJ.log("inverted: "+ip2.getStatistics());
new ImagePlus("inverted",ip2).show();

>> -----Original Message-----
>> From: ImageJ Interest Group [mailto:[hidden email]] On Behalf Of
>> Gabriel Landini
>> Sent: Tuesday, June 02, 2015 10:54 AM
>> To: [hidden email]
>> Subject: Re: Clearing 32 bit float images
>>
>> On Tuesday 02 Jun 2015 10:09:55 Burger Wilhelm wrote:
>>> when applying 'Edit -> Clear' on a 32 bit float image (FloatProcessor) I
>>> would assume that the affected pixel values are set to zero. However, the
>>> command inserts some other (display-related?) value. E.g., on the "bridge"
>>> sample image, the new value is 85.0.
>>>
>>> Is this intended behaviour? If yes, how is the clear value determined?
>>
>> Hi Wilhelm,
>> It uses the values in the colour picker.
>> If you set the background colour to 0 it clears with that value.
>>
>> Cheers
>>
>> Gabriel

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

Burger Wilhelm

Re: Clearing 32 bit float images

In reply to this post by Michael Schmid

Makes sense, I was not aware of this 'convention'.
Thanks to Micheal and Wayne for clarifying this!

--Wilhelm

________________________________________
From: ImageJ Interest Group [[hidden email]] On Behalf Of Michael Schmid [[hidden email]]
Sent: Tuesday, June 02, 2015 14:41
To: [hidden email]
Subject: Re: Clearing 32 bit float images

Hi Wilhelm,

for setting a pixel value to zero (or any other [uncalibrated] numeric value): Use Process>Math>Set.
If you have a float image and you want to invert it by changing the sign, use Process>Math>Multiply to multiply by -1.

The Edit>Fill (Edit>Clear) command uses the eyedropper foreground (background) color or whatever comes closest for the way the image is currently displayed.

Edit>Invert uses the current display range, i.e. the min and max that you can set in the Brightness&Contrast panel. So you can have e.g. a 16-bit image, display it as 12-bit image (0-4095) and have 'Invert' create the correct values if if you would invert a 12-bit image (0 becomes 4095, and 4095 becomes 0). You could also have a float image with pixel values 0-255, display it with this range, and do the 'Invert' operation just as if it were an 8-bit image, but with floating-point accuracy.

With this convention, Edit>Invert also ensures that an image that is displayable with the current B&C settings remains displayable after 'Invert'. If you have a floating-point image with pixels e.g. 0-255, and you would invert it by multiplying with -1, it will be all black when shown with the 0-255 display range.

Michael
________________________________________________________________
On Jun 2, 2015, at 11:26, Burger Wilhelm wrote:

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