Difference between revisions of "Build streaming capabilities for output image generation"

From OTBWiki
Jump to: navigation, search
m (Reverted edits by FernandoMorani (Talk) to last revision by AnnaSharp)
(Removed spam)
 
Line 15: Line 15:
 
=== Conclusion ===
 
=== Conclusion ===
  
A simple pipeline with a streaming writer at the end will not do it. We have to design a specific solution to make this process streamable. The first step would be to know if we can have an idea or set arbitrary bounds. [http://uk.bestessays.com uk essays]
+
A simple pipeline with a streaming writer at the end will not do it. We have to design a specific solution to make this process streamable. The first step would be to know if we can have an idea or set arbitrary bounds.
  
 
If we consider the maximum deformation as an input parameter for both algorithms, we can wrap those filters to give them streaming capabilities, by padding the requested region with the maximal deformation. This will require two additional filters to be implemented, wrapping the [http://www.orfeo-toolbox.org/doxygen/classitk_1_1PDEDeformableRegistrationFilter.html itk::PDEDeformableRegistrationFilter] and [http://www.orfeo-toolbox.org/doxygen/classitk_1_1WarpImageFilter.html itk::WarpImageFilter]. These filters will be used in the fine registration model in order to be able to generate the deformation field and the warped image for a large input.
 
If we consider the maximum deformation as an input parameter for both algorithms, we can wrap those filters to give them streaming capabilities, by padding the requested region with the maximal deformation. This will require two additional filters to be implemented, wrapping the [http://www.orfeo-toolbox.org/doxygen/classitk_1_1PDEDeformableRegistrationFilter.html itk::PDEDeformableRegistrationFilter] and [http://www.orfeo-toolbox.org/doxygen/classitk_1_1WarpImageFilter.html itk::WarpImageFilter]. These filters will be used in the fine registration model in order to be able to generate the deformation field and the warped image for a large input.

Latest revision as of 15:58, 27 August 2013

Registration and deformation field estimation step

The registration filter used in the FineRegistrationApplication inherit from itk::PDEDeformableRegistrationFilter.

This class contains a GenerateInputRequestedRegion() which is requesting the output requested region for the fixed input and the largest possible region for the moving input.

As a result, this filter can not be streamed straightforward.

Moving image warping step

the itk::WarpImageFilter is used to warp the moving image according to the deformation field. It contains a GenerateInputRequestedRegion() which is requesting the output requested region for the input deformation field and the largest possible region for the input image to warp.

As a result, this filter can not be streamed straightforward.

Conclusion

A simple pipeline with a streaming writer at the end will not do it. We have to design a specific solution to make this process streamable. The first step would be to know if we can have an idea or set arbitrary bounds.

If we consider the maximum deformation as an input parameter for both algorithms, we can wrap those filters to give them streaming capabilities, by padding the requested region with the maximal deformation. This will require two additional filters to be implemented, wrapping the itk::PDEDeformableRegistrationFilter and itk::WarpImageFilter. These filters will be used in the fine registration model in order to be able to generate the deformation field and the warped image for a large input.