| neher | |
| Jan 30 2015, 12:06 PM |
| F1223: pic3dRefLinear_Data.png | |
| Oct 7 2015, 4:45 PM |
| F1222: Pic3dRef_Linear_neuRausgeschrieben.png | |
| Oct 7 2015, 4:44 PM |
| F1221: Pic3dRefCubic_neuRausgeschrieben.png | |
| Oct 7 2015, 4:40 PM |
| F1220: pic3dRefCubic_Data.png | |
| Oct 7 2015, 4:40 PM |
Current code:
return this->GetIndexToWorldTransform()->TransformPoint(m_BoundingBox->GetCenter());
Proposed:
if (m_ImageGeometry)
{
Point3D c = m_BoundingBox->GetCenter();
c[0] -= 0.5;
c[1] -= 0.5;
c[2] -= 0.5;
IndexToWorld(c, c);
return c;
}
This proposed code seems to be right, but there are a lot of work-arounds in other classes (e.g. BoundingObject.cpp for image cropper), which need to be found and fixed.
These tests are still failing - it might be wrong testing or wrong testing data. Has to be checked!
mitkImageVtkMapper2D_pic3dSwivel640x480 (Failed)
ResliceInterpolationIsLinear (Failed)
ResliceInterpolationIsCubic (Failed)
User wildes has pushed new remote branch:
bug-18695-AdaptGetCenterForImageGeometries_IntBranch
Core Change Request
GetCenter() returned wrong center of images with m_ImageGeometry == true (center based geometries).
Center was shifted by 0.5.
Return the correct center, no matter if the object is center or corner based.
No check for ImageGeometry variable.
Shift the center by 0.5.
mitkBaseGeometry - correct the GetCenter method
mitkBoundingObject - here was a workaround for the problem
mitkGeometry3DTest - test did not cover the center check for center based images
Testing data for the following tests needed to be changed, as the rotation center was wrong:
mitkImageVtkMapper2D_pic3dSwivel640x480
ResliceInterpolationIsLinear
ResliceInterpolationIsCubic
Geometry3DTest - new if case in GetCenter Test method.
[85d7b9]: Merge branch 'bug-18695-AdaptGetCenterForImageGeometries_IntBranch'
2015-09-30 15:30:58 Esther Wild [a7cf4c]
Create new MITK testing data
Images were rotated around wrong center, due to bug in GetCenter-method.
2015-09-30 13:23:21 Esther Wild [e03152]
Merge branch 'bug-18695-AdaptGetCenterForImageGeometries' into bug-18695-AdaptGetCenterForImageGeometries_IntBranch
2015-09-09 17:35:35 Esther Wild [dbfc57]
Correct GetCenter for image geometries.
[d88bef]: Merge branch 'bug-18695-AdaptGetCenterForImageGeometries_IntBranch'
2015-10-05 11:32:06 Esther Wild [94dd70]
COMP: tests deactivated, because win 32 bit continuous renders images slightly different.
Rendering test fails on 32bin Win continuous. Deactivated. ToDo: Find out, why the rendering is different on 32 bit.
The rendering on the windows continuous is different and has always been different. However, with the old data, the differences were small enough to be below the threshold of 0.35. With the new data, the threshold needs to be something between 1 and 5...
Adapting the threshold is not the right solution, it would be better to add a second testing image.
Differences are not caused by 32 bit system. It might be the QT version (continuous qt5.3.0, local windows qt5.4.0), or the rendering in a virtual machine.
Spacing and Geometries (currentWorldPlaneGeometry and CurrentWOrldGeometry) are identical on both machines, therefore, it should be a rendering problem.
The image of the virtual machine looks like an interpolation filter was applied. However, we did not find such a difference yet.
Reference image of the ResliceInterpolationIsCubic test, passing on local machine or linux.
Reference image of the ResliceInterpolationIsCubic test, passing on the Windows continuous dartclient.
Reference image of the ResliceInterpolationIsLinear test, passing on Windows continuous.
Reference image of the ResliceInterpolationIsLinear test, passing on linux or windows (local).
