RGB to binary problem

[Tal Shprung]

Dear colleagues,

I have tried to change RGB images to binary by going to process -> binary ->
make binary, but in this way I get a lot of background noise in many of the
images.

By chance I have done the process going through 32-bit gray scale (image ->
type -> 32-bit) and than "make binary" and saw that it bypasses the
background noise problem.

Have anyone else encountered this phenomenon? Are there any comments?

Thanks,

Tal

[Albert Cardona]

> I have tried to change RGB images to binary by going to process -> binary ->
> make binary, but in this way I get a lot of background noise in many of the
> images.
>
> By chance I have done the process going through 32-bit gray scale (image ->
> type -> 32-bit) and than "make binary" and saw that it bypasses the
> background noise problem.
>
> Have anyone else encountered this phenomenon? Are there any comments?
>  

I suspect your RGB image goes to 8-bit color first (which is terrible
for any quantification, since pixel intensities no longer correlate to
color intensity), and then to binary.

RGB to 32 bit will do the right thing: create a luminance image and put
it into a float array.

You can see what RGB really is if you convert the RGB to HSB, and then
use the 'B' slice of the 3-slice stack (the Brightness channel) which is
essentially the grayscale version of your image (Hue and Saturation
containing the color information.)

Albert

--
Albert Cardona
http://www.mcdb.ucla.edu/Research/Hartenstein/acardona

[Tal Shprung]

Wayne Rasband told me to do RGB split and then use the red channel (as my
pictures are with only red color).

I have tried it and saw that it gives the cleanest binary picture (compared
to going straight to binary or even to 32bit or 8bit grayscale and then
binary).
Thank you all.

Tal

On Tue, Jun 24, 2008 at 9:02 PM, Albert Cardona <[hidden email]>
wrote:

>  I have tried to change RGB images to binary by going to process -> binary
>> ->
>> make binary, but in this way I get a lot of background noise in many of
>> the
>> images.
>>
>> By chance I have done the process going through 32-bit gray scale (image
>> ->
>> type -> 32-bit) and than "make binary" and saw that it bypasses the
>> background noise problem.
>>
>> Have anyone else encountered this phenomenon? Are there any comments?
>>
>>
>
>
> I suspect your RGB image goes to 8-bit color first (which is terrible for
> any quantification, since pixel intensities no longer correlate to color
> intensity), and then to binary.
>
> RGB to 32 bit will do the right thing: create a luminance image and put it
> into a float array.
>
> You can see what RGB really is if you convert the RGB to HSB, and then use
> the 'B' slice of the 3-slice stack (the Brightness channel) which is
> essentially the grayscale version of your image (Hue and Saturation
> containing the color information.)
>
> Albert
>
> --
> Albert Cardona
> http://www.mcdb.ucla.edu/Research/Hartenstein/acardona
>

[Doug S-3]

Tal or Wayne,
    about using just the red channel,
is that specifically for the type of images you are using
( e.g they have better contrast in that channel )
or is there something inherently better about the red channel
as opposed to the blue or green?
Doug

Tal Shprung wrote:

> Wayne Rasband told me to do RGB split and then use the red channel (as my
> pictures are with only red color).
>
> I have tried it and saw that it gives the cleanest binary picture (compared
> to going straight to binary or even to 32bit or 8bit grayscale and then
> binary).
> Thank you all.
>
> Tal
>
> On Tue, Jun 24, 2008 at 9:02 PM, Albert Cardona <[hidden email]>
> wrote:
>
>  
>>  I have tried to change RGB images to binary by going to process -> binary
>>    
>>> ->
>>> make binary, but in this way I get a lot of background noise in many of
>>> the
>>> images.
>>>
>>> By chance I have done the process going through 32-bit gray scale (image
>>> ->
>>> type -> 32-bit) and than "make binary" and saw that it bypasses the
>>> background noise problem.
>>>
>>> Have anyone else encountered this phenomenon? Are there any comments?
>>>
>>>
>>>      
>> I suspect your RGB image goes to 8-bit color first (which is terrible for
>> any quantification, since pixel intensities no longer correlate to color
>> intensity), and then to binary.
>>
>> RGB to 32 bit will do the right thing: create a luminance image and put it
>> into a float array.
>>
>> You can see what RGB really is if you convert the RGB to HSB, and then use
>> the 'B' slice of the 3-slice stack (the Brightness channel) which is
>> essentially the grayscale version of your image (Hue and Saturation
>> containing the color information.)
>>
>> Albert
>>
>> --
>> Albert Cardona
>> http://www.mcdb.ucla.edu/Research/Hartenstein/acardona
>>
>>    
>
>  

[Tal Shprung]

I am using only the red channel in my images. I guess that although the blue
and green channels are empty, when converting directly to binary they make
some noise  for some reason.

On Tue, Jun 24, 2008 at 9:44 PM, Doug S <[hidden email]> wrote:

> Tal or Wayne,
>   about using just the red channel,
> is that specifically for the type of images you are using
> ( e.g they have better contrast in that channel )
> or is there something inherently better about the red channel
> as opposed to the blue or green?
> Doug
>
>
> Tal Shprung wrote:
>
>> Wayne Rasband told me to do RGB split and then use the red channel (as my
>> pictures are with only red color).
>>
>> I have tried it and saw that it gives the cleanest binary picture
>> (compared
>> to going straight to binary or even to 32bit or 8bit grayscale and then
>> binary).
>> Thank you all.
>>
>> Tal
>>
>> On Tue, Jun 24, 2008 at 9:02 PM, Albert Cardona <
>> [hidden email]>
>> wrote:
>>
>>
>>
>>>  I have tried to change RGB images to binary by going to process ->
>>> binary
>>>
>>>
>>>> ->
>>>> make binary, but in this way I get a lot of background noise in many of
>>>> the
>>>> images.
>>>>
>>>> By chance I have done the process going through 32-bit gray scale (image
>>>> ->
>>>> type -> 32-bit) and than "make binary" and saw that it bypasses the
>>>> background noise problem.
>>>>
>>>> Have anyone else encountered this phenomenon? Are there any comments?
>>>>
>>>>
>>>>
>>>>
>>> I suspect your RGB image goes to 8-bit color first (which is terrible for
>>> any quantification, since pixel intensities no longer correlate to color
>>> intensity), and then to binary.
>>>
>>> RGB to 32 bit will do the right thing: create a luminance image and put
>>> it
>>> into a float array.
>>>
>>> You can see what RGB really is if you convert the RGB to HSB, and then
>>> use
>>> the 'B' slice of the 3-slice stack (the Brightness channel) which is
>>> essentially the grayscale version of your image (Hue and Saturation
>>> containing the color information.)
>>>
>>> Albert
>>>
>>> --
>>> Albert Cardona
>>> http://www.mcdb.ucla.edu/Research/Hartenstein/acardona
>>>
>>>
>>>
>>
>>
>>
>