Re: was colour bug? - pixel data types

[Daniel James White]

Hi Mike,

On Mar 29, 2011, at 6:00 AM, IMAGEJ automatic digest system wrote:

>
> Date:    Mon, 28 Mar 2011 11:43:14 +0100
> From:    Mike A <[hidden email]>
> Subject: colour bug?
>
> Hi folks,
>
> I am using imageJ v1.45e11
>
> I am opening a two channel 2D stack (x,y) stack with LOCI bioformats and colour mode set to “composite”.  If I save this as .tif I get a 16bit RGB image.  great !

Jolly good.

>
> However,  If i do image > colour > split channels, and save one of the channels as .tif, it comes out in greyscale in photoshop.

That is what should happen. Its not as bug, its the expected behaviour!

>
> Importantly, if I open my stack with LOCI with colour mode set to “default” and the split channels function checked my channels split and appear greyscale.  If i use image > colour > merge channels and select a single one of my split channels in red and “none” for the other channels with “create composite” checked nothing happens.  Basically it is not possible to save a single channel in my stack in 16bit colour.  Is this a bug?
>
> I presumed that I could use the merge channels option to place the channel that im interested in, in lets say, the red channel, and select none for blue/green/grey with create composite checked, to output a 16bit RGB colour file in which the channel im interested in appeared red, and blue/green were present in the 16bit rgb image but just contained zeros/no data, thus allowing me to save the channel im interested in in 16bit colour.

Don't worry, you are just missing the basics of how different pixel types work in imageJ and elsewhere.

You can see the different pixel data types in menu command
Image - Type

The simples types are "greyscale" :
8 - bit, 16-bit and 32 bit.
There is only once value per pixel.
8 bit holds INTEGER ONLY values in the range 0-255  (2 to the power 8 different intensity levels"
16 bit holds INTEGER ONLY  values from 0 to ( (2 to the power 16) -1 ) about 65000 and some.
32 bit is a floating point representation where you can have numbers in a huge range with a decimal point somewhere in the number.

If you have a multi channel image with n channels... there is effectively n pixel values for each pixel.
ImageJ composite image representation shows this as a channel slider a the bottom of the window.
A composite image can have channels with 8, 16 or 32 bit pixels.

Another way to show a multi channel image is to use RGB pixel type,
where there are exactly 3 values for each pixel, one for each red green and blue "channels"
BUT, RGB pixel type is 8 bit... so if you started with nice data from a 12 bit or 16 bit camera,
if you convert to RGB, then you will lose a great deal of intensity resolution (which matters only for high signal:noise images)

The imageJ way to do what you want is to get the single channel image,
use one of the display look up tables (the LUT button in the toolbar)
to choose how you would like it to be shown in colour,
then change the pixel type to RGB (or maybe 8 bit color)
then save that as a tiff.

There is no need to use photoshop for annotation of the images, as you can do all that in imageJ.

For figure layout, photoshop is not well suited,
and something like illustrator is better suited... or some other open source document layout software.

In fact I dont know why we have to do any layout, apart from a schematic one of what goes where,
since the publishers have to totally re do the layout for publication anyway...
reviewers need PDF, but the publishers should make that (or else what are they actually for?)

Actually the reviewers need the raw image data .... not some tiny compressed and smashed image in an PDF.
They must be able to get the raw image and repeat your analysis from it.

cheers

Dan



>
> Cheers
>
> Mike

Dr. Daniel James White BSc. (Hons.) PhD
Senior Microscopist / Image Visualisation, Processing and Analysis
Light Microscopy and Image Processing Facilities
Max Planck Institute of Molecular Cell Biology and Genetics
Pfotenhauerstrasse 108
01307 DRESDEN
Germany

[phone] (German Mobile)
[phone] (Work phone at MPI-CBG)
[phone] (Fax MPI-CBG LMF)

http://www.bioimagexd.net  BioImageXD
http://pacific.mpi-cbg.de                Fiji -  is just ImageJ (Batteries Included)
http://www.chalkie.org.uk                Dan's Homepages
https://ifn.mpi-cbg.de  Dresden Imaging Facility Network
dan (at) chalkie.org.uk
( white (at) mpi-cbg.de )

[Mike A]

Daniel, thanks very much for your reply.  Very informative. I tried your
method:

"The imageJ way to do what you want is to get the single channel image,
use one of the display look up tables (the LUT button in the toolbar)
to choose how you would like it to be shown in colour,
then change the pixel type to RGB (or maybe 8 bit color)
then save that as a tiff."

However, when i do this it saves the tiff as an 8bit RGB and not a 16bit
RGB.  This happens on my stack, and also on sample stacks.  for example:
1. open organ of corti
2. image > color > split channels
3. change channel 4 to lets say magenta by clicking the "primary colours"
button on the "lookup tables" toolbar
4. change image > type from 16bit to RGB color
5. save as tiff

The resulting tiff is an 8bit RGB image and not a 16bit RGB image.  Am i
still missing something?

Thanks for your help.

Mike A


-----Original Message-----
From: Daniel James White
Sent: Tuesday, March 29, 2011 9:04 AM
To: ImageJ Interest Group ; [hidden email]
Subject: Re: was colour bug? - pixel data types

Hi Mike,

On Mar 29, 2011, at 6:00 AM, IMAGEJ automatic digest system wrote:

>
> Date:    Mon, 28 Mar 2011 11:43:14 +0100
> From:    Mike A <[hidden email]>
> Subject: colour bug?
>
> Hi folks,
>
> I am using imageJ v1.45e11
>
> I am opening a two channel 2D stack (x,y) stack with LOCI bioformats and
> colour mode set to “composite”.  If I save this as .tif I get a 16bit RGB
> image.  great !

Jolly good.

>
> However,  If i do image > colour > split channels, and save one of the
> channels as .tif, it comes out in greyscale in photoshop.

That is what should happen. Its not as bug, its the expected behaviour!

>
> Importantly, if I open my stack with LOCI with colour mode set to
>  “default” and the split channels function checked my channels split and
> appear greyscale.  If i use image > colour > merge channels and select a
> single one of my split channels in red and “none” for the other channels
> with “create composite” checked nothing happens.  Basically it is not
> possible to save a single channel in my stack in 16bit colour.  Is this a
> bug?
>
> I presumed that I could use the merge channels option to place the channel
> that im interested in, in lets say, the red channel, and select none for
> blue/green/grey with create composite checked, to output a 16bit RGB
> colour file in which the channel im interested in appeared red, and
> blue/green were present in the 16bit rgb image but just contained zeros/no
> data, thus allowing me to save the channel im interested in in 16bit
> colour.

Don't worry, you are just missing the basics of how different pixel types
work in imageJ and elsewhere.

You can see the different pixel data types in menu command
Image - Type

The simples types are "greyscale" :
8 - bit, 16-bit and 32 bit.
There is only once value per pixel.
8 bit holds INTEGER ONLY values in the range 0-255  (2 to the power 8
different intensity levels"
16 bit holds INTEGER ONLY  values from 0 to ( (2 to the power 16) -1 ) about
65000 and some.
32 bit is a floating point representation where you can have numbers in a
huge range with a decimal point somewhere in the number.

If you have a multi channel image with n channels... there is effectively n
pixel values for each pixel.
ImageJ composite image representation shows this as a channel slider a the
bottom of the window.
A composite image can have channels with 8, 16 or 32 bit pixels.

Another way to show a multi channel image is to use RGB pixel type,
where there are exactly 3 values for each pixel, one for each red green and
blue "channels"
BUT, RGB pixel type is 8 bit... so if you started with nice data from a 12
bit or 16 bit camera,
if you convert to RGB, then you will lose a great deal of intensity
resolution (which matters only for high signal:noise images)

The imageJ way to do what you want is to get the single channel image,
use one of the display look up tables (the LUT button in the toolbar)
to choose how you would like it to be shown in colour,
then change the pixel type to RGB (or maybe 8 bit color)
then save that as a tiff.

There is no need to use photoshop for annotation of the images, as you can
do all that in imageJ.

For figure layout, photoshop is not well suited,
and something like illustrator is better suited... or some other open source
document layout software.

In fact I dont know why we have to do any layout, apart from a schematic one
of what goes where,
since the publishers have to totally re do the layout for publication
anyway...
reviewers need PDF, but the publishers should make that (or else what are
they actually for?)

Actually the reviewers need the raw image data .... not some tiny compressed
and smashed image in an PDF.
They must be able to get the raw image and repeat your analysis from it.

cheers

Dan



>
> Cheers
>
> Mike

Dr. Daniel James White BSc. (Hons.) PhD
Senior Microscopist / Image Visualisation, Processing and Analysis
Light Microscopy and Image Processing Facilities
Max Planck Institute of Molecular Cell Biology and Genetics
Pfotenhauerstrasse 108
01307 DRESDEN
Germany

[phone] (German Mobile)
[phone] (Work phone at MPI-CBG)
[phone] (Fax MPI-CBG LMF)

http://www.bioimagexd.net BioImageXD
http://pacific.mpi-cbg.de Fiji -  is just ImageJ (Batteries Included)
http://www.chalkie.org.uk Dan's Homepages
https://ifn.mpi-cbg.de Dresden Imaging Facility Network
dan (at) chalkie.org.uk
( white (at) mpi-cbg.de )

[Michael Schmid]

Hi Mike,

Image type "RGB color" in ImageJ has 8 bits per color, so your type  
conversion in step 4 to RGB color changes it to 8 bits.
If you want to keep 16 bits/channel, you have to keep it as composite  
image.

Other programs won't correctly interpret composite images saved as  
TIFF unless they have only 3 colors, R, G, B.

To get it as 3*16 bits, you would have to do some arithmetics to  
calculate what the R, B, G values are (with 16-bit accuracy). For the  
Organs of Corti as displayed usually (R, G, B + gray), duplicate the  
4th channel (gray) and add it to the R, G, B channels. Then delete  
the gray channel.
If you want the 4th channel to appear as cyan, it should add to the R  
and B channels only.

Hope this helps,

Michael
________________________________________________________________

On 29 Mar 2011, at 10:47, Mike A wrote:

> Daniel, thanks very much for your reply.  Very informative. I tried  
> your method:
>
> "The imageJ way to do what you want is to get the single channel  
> image,
> use one of the display look up tables (the LUT button in the toolbar)
> to choose how you would like it to be shown in colour,
> then change the pixel type to RGB (or maybe 8 bit color)
> then save that as a tiff."
>
> However, when i do this it saves the tiff as an 8bit RGB and not a  
> 16bit RGB.  This happens on my stack, and also on sample stacks.  
> for example:
> 1. open organ of corti
> 2. image > color > split channels
> 3. change channel 4 to lets say magenta by clicking the "primary  
> colours" button on the "lookup tables" toolbar
> 4. change image > type from 16bit to RGB color
> 5. save as tiff
>
> The resulting tiff is an 8bit RGB image and not a 16bit RGB image.  
> Am i still missing something?
>
> Thanks for your help.
>
> Mike A
>
>
> -----Original Message----- From: Daniel James White
> Sent: Tuesday, March 29, 2011 9:04 AM
> To: ImageJ Interest Group ; [hidden email]
> Subject: Re: was colour bug? - pixel data types
>
> Hi Mike,
>
> On Mar 29, 2011, at 6:00 AM, IMAGEJ automatic digest system wrote:
>
>>
>> Date:    Mon, 28 Mar 2011 11:43:14 +0100
>> From:    Mike A <[hidden email]>
>> Subject: colour bug?
>>
>> Hi folks,
>>
>> I am using imageJ v1.45e11
>>
>> I am opening a two channel 2D stack (x,y) stack with LOCI  
>> bioformats and colour mode set to “composite”.  If I save this  
>> as .tif I get a 16bit RGB image.  great !
>
> Jolly good.
>
>>
>> However,  If i do image > colour > split channels, and save one of  
>> the channels as .tif, it comes out in greyscale in photoshop.
>
> That is what should happen. Its not as bug, its the expected  
> behaviour!
>
>>
>> Importantly, if I open my stack with LOCI with colour mode set to  
>> “default” and the split channels function checked my channels  
>> split and appear greyscale.  If i use image > colour > merge  
>> channels and select a single one of my split channels in red and  
>> “none” for the other channels with “create composite” checked  
>> nothing happens.  Basically it is not possible to save a single  
>> channel in my stack in 16bit colour.  Is this a bug?
>>
>> I presumed that I could use the merge channels option to place the  
>> channel that im interested in, in lets say, the red channel, and  
>> select none for blue/green/grey with create composite checked, to  
>> output a 16bit RGB colour file in which the channel im interested  
>> in appeared red, and blue/green were present in the 16bit rgb  
>> image but just contained zeros/no data, thus allowing me to save  
>> the channel im interested in in 16bit colour.
>
> Don't worry, you are just missing the basics of how different pixel  
> types work in imageJ and elsewhere.
>
> You can see the different pixel data types in menu command
> Image - Type
>
> The simples types are "greyscale" :
> 8 - bit, 16-bit and 32 bit.
> There is only once value per pixel.
> 8 bit holds INTEGER ONLY values in the range 0-255  (2 to the power  
> 8 different intensity levels"
> 16 bit holds INTEGER ONLY  values from 0 to ( (2 to the power 16)  
> -1 ) about 65000 and some.
> 32 bit is a floating point representation where you can have  
> numbers in a huge range with a decimal point somewhere in the number.
>
> If you have a multi channel image with n channels... there is  
> effectively n pixel values for each pixel.
> ImageJ composite image representation shows this as a channel  
> slider a the bottom of the window.
> A composite image can have channels with 8, 16 or 32 bit pixels.
>
> Another way to show a multi channel image is to use RGB pixel type,
> where there are exactly 3 values for each pixel, one for each red  
> green and blue "channels"
> BUT, RGB pixel type is 8 bit... so if you started with nice data  
> from a 12 bit or 16 bit camera,
> if you convert to RGB, then you will lose a great deal of intensity  
> resolution (which matters only for high signal:noise images)
>
> The imageJ way to do what you want is to get the single channel image,
> use one of the display look up tables (the LUT button in the toolbar)
> to choose how you would like it to be shown in colour,
> then change the pixel type to RGB (or maybe 8 bit color)
> then save that as a tiff.
>
> There is no need to use photoshop for annotation of the images, as  
> you can do all that in imageJ.
>
> For figure layout, photoshop is not well suited,
> and something like illustrator is better suited... or some other  
> open source document layout software.
>
> In fact I dont know why we have to do any layout, apart from a  
> schematic one of what goes where,
> since the publishers have to totally re do the layout for  
> publication anyway...
> reviewers need PDF, but the publishers should make that (or else  
> what are they actually for?)
>
> Actually the reviewers need the raw image data .... not some tiny  
> compressed and smashed image in an PDF.
> They must be able to get the raw image and repeat your analysis  
> from it.
>
> cheers
>
> Dan
>
>
>
>>
>> Cheers
>>
>> Mike
>
> Dr. Daniel James White BSc. (Hons.) PhD
> Senior Microscopist / Image Visualisation, Processing and Analysis
> Light Microscopy and Image Processing Facilities
> Max Planck Institute of Molecular Cell Biology and Genetics
> Pfotenhauerstrasse 108
> 01307 DRESDEN
> Germany
>
> [phone] (German Mobile)
> [phone] (Work phone at MPI-CBG)
> [phone] (Fax MPI-CBG LMF)
>
> http://www.bioimagexd.net BioImageXD
> http://pacific.mpi-cbg.de Fiji -  is just ImageJ (Batteries Included)
> http://www.chalkie.org.uk Dan's Homepages
> https://ifn.mpi-cbg.de Dresden Imaging Facility Network
> dan (at) chalkie.org.uk
> ( white (at) mpi-cbg.de )