ITK/Release 4/Refactor Numerical Libraries/Inventory/Matrix Vector Operations: Difference between revisions
From KitwarePublic
< ITK | Release 4 | Refactor Numerical Libraries | Inventory
Jump to navigationJump to search
| Line 7: | Line 7: | ||
These methods perform Matrix x Vector multiplications. | These methods perform Matrix x Vector multiplications. | ||
* The first one converts the integer grid coordinates of a pixel to it s world coordinates in physical space. | |||
* The second method converts th world coordinates in physical space to integer grid coordinates of a pixel. | |||
Internally they use the helper matrices | |||
this->m_IndexToPhysicalPoint | |||
this->m_PhysicalPointToIndex | |||
That are pre-computed in the method: | |||
ComputeIndexToPhysicalPointMatrices() | |||
in lines 178-201 if itkImageBase.hxx | |||
The method | The method | ||
| Line 19: | Line 34: | ||
in lines: | in lines: | ||
456 - 480 | |||
Luis to find an example suitable for profiling. Probably based on the ResampleImage*.cxx files in | |||
ITK/Examples/Filtering | |||
but with a 3D input image, so it takes enough computation time to provide reliable benchmarking. | |||
Revision as of 15:41, 16 September 2011
Index to Point
The ImageBase class provides the methods:
TransformIndexToPhysicalPoint TransformPhysicalPointToIndex
These methods perform Matrix x Vector multiplications.
- The first one converts the integer grid coordinates of a pixel to it s world coordinates in physical space.
- The second method converts th world coordinates in physical space to integer grid coordinates of a pixel.
Internally they use the helper matrices
this->m_IndexToPhysicalPoint this->m_PhysicalPointToIndex
That are pre-computed in the method:
ComputeIndexToPhysicalPointMatrices()
in lines 178-201 if itkImageBase.hxx
The method
TransformIndexToPhysicalPoint
Is called millions of times during the execution of Image Registration processes.
The actual multiplication is done in the file
ITK/Core/Common/include/itkImageBase.h
in lines:
456 - 480
Luis to find an example suitable for profiling. Probably based on the ResampleImage*.cxx files in
ITK/Examples/Filtering
but with a 3D input image, so it takes enough computation time to provide reliable benchmarking.
