[Insight-users] Filters Input Image Pixel Type

[Mauro Maiorca]

glad it helped!

almost all the examples cast the input type to internal type (float or
double) for processing, and then they cast the internal image to an output
type. It's more accurate to work with floating point types for
processing, especially with filters that do differential computations,
smoothing, optimization, etc. etc.
Obviously be sure you don't exceed the numeric_limits of a data type, for
example if for some reasons you have to cast a CT image with values in HU
(in a range [-1200, 3000]) to an unsigned char type (which is in a range
[1,255]), you might want to stretch the intensity values beforehand.

Have fun!

ciao,
Mauro


On Thu, Dec 12, 2013 at 10:35 PM, elena bresciani <
elena.bresciani87 at gmail.com> wrote:

> Ok ok, I get it! :)
>
> Thank you very much
>
> Cheers,
> Elena
>
>
>
> 2013/12/12 Mauro Maiorca <mauromaiorca at gmail.com>
>
>> On Thu, Dec 12, 2013 at 9:06 PM, elena bresciani <
>> elena.bresciani87 at gmail.com> wrote:
>>
>>> > Ok, thanks but my question was more like How can I know which input
>> image type requires a particular filter?
>>
>> ah okay! Any descent of ImageToImageFilter it is meant to support at
>> least all the native types such as float, char, int, and so on. As all the
>> classes are templated, so the numeric types are those compatible with your
>> C++ compiler. However, when you write an image on a file you are restricted
>> on the data types supported by the image format you are using, for example
>> MRCImageIO complains with scalar double types, while double type
>> works perfectly fine with MetaIO format.
>>
>> > For example: itk::GradientMagnitudeImageFilter requires the input image
>> to have a float pixel type (as described in the software guide).
>> > Where can I find this kind of informations?
>>
>> In the case of itk::GradientMagnitudeImageFilter you might want to have
>> a floating point pixel values, as you might want a continuous histogram,
>> i.e. you don't want to approximate real values to an integer type, as you
>> loose all the decimals. But itk::GradientMagnitudeImageFilter works
>> perfectly fine with integers too. Another example, if you want to threshold
>> an image, it's unnecessary to use a floating point pixel type, as usually
>> you need to store integer values, not decimals. But it works perfectly fine
>> using floating pixel types. It really depends on what is the problem you
>> are going to solve. Does it make sense?
>>
>> cheers,
>> Mauro
>>
>>
>>
>>
>>
>> 2013/12/12 Mauro Maiorca <mauromaiorca at gmail.com>
>>>
>>>> Hi Elena,
>>>>
>>>> have a look at the ITK software guide (
>>>> http://www.itk.org/ItkSoftwareGuide.pdf), chapter 4 (User's guide ->
>>>> Data Representation -> Image), it might answer your question.
>>>>
>>>> Section 4.1.1 shows how to create a 3D image with unsigned short pixel
>>>> data, in summary you have to define your image type as:
>>>> typedef itk::Image< unsigned short, 3 > ImageType;
>>>> If you want to create an image with unsigned int pixel data, you have
>>>> to replace "unsigned short" with "unsigned int", i.e.
>>>> typedef itk::Image< unsigned int, 3 > ImageType;
>>>>
>>>> You can play with the examples that you can find either in your ITK
>>>> source directory (Examples/DataRepresentation/Image/Image1.cxx, etc etc) or
>>>> online (
>>>> https://github.com/Kitware/ITK/blob/master/Examples/DataRepresentation/Image/Image1.cxxetc etc)
>>>>
>>>> If you want to cast image types (for example converting from unsigned
>>>> short to unsigned int), have a look at section "6.3 Casting and Intensity
>>>> Mapping" of the ITK software guide.
>>>>
>>>> Hope it helps.
>>>>
>>>> cheers,
>>>> Mauro
>>>>
>>>>
>>>>
>>>>
>>>>
>>>> On Thu, Dec 12, 2013 at 8:19 PM, elena bresciani <
>>>> elena.bresciani87 at gmail.com> wrote:
>>>>
>>>>> Hi to everybody,
>>>>>
>>>>> I have a (probably really silly) question about which kind of input
>>>>> image pixel type does a filter require.
>>>>> Is it wrote in the doxygen documentation? If yes, where?
>>>>>
>>>>> I'm obviously a beginner and I can't understand well how to read the
>>>>> documentation...
>>>>>
>>>>> Thank you in advance,
>>>>>
>>>>> Elena
>>>>>
>>>>>
>>>>>
>>>>> _____________________________________
>>>>> Powered by www.kitware.com
>>>>>
>>>>> Visit other Kitware open-source projects at
>>>>> http://www.kitware.com/opensource/opensource.html
>>>>>
>>>>> Kitware offers ITK Training Courses, for more information visit:
>>>>> http://www.kitware.com/products/protraining.php
>>>>>
>>>>> Please keep messages on-topic and check the ITK FAQ at:
>>>>> http://www.itk.org/Wiki/ITK_FAQ
>>>>>
>>>>> Follow this link to subscribe/unsubscribe:
>>>>> http://www.itk.org/mailman/listinfo/insight-users
>>>>>
>>>>>
>>>>
>>>
>>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.itk.org/pipermail/insight-users/attachments/20131212/029b25a7/attachment.htm>