[vtkusers] ComputeTransformedOrigin bug in vtkImageReader.cxx

[Josh Snyder]

Making this change fixed my problem. Could someone verify this and check 
it in?

Josh

Josh Snyder wrote:

> I've run into a problem using vtkImageReader with the SetTransform 
> functionality, and I think it's due to the line of code referenced in 
> the post below (line 495 in the current CVS version). It seems to me 
> that the older version (w/out the +1) was correct. My understanding is 
> that if the Transform specifies an axis reversal, the order of the 
> voxels along that axis needs to be reversed, and the origin coordinate 
> for that axis needs to be shifted so that the voxels originally 
> occurring at negative locations become positive and vice versa. The 
> former step is handled by the inverse transform of the increments 
> during the vtkImageReaderUpdate2 method, and the latter by the code 
> mentioned below. I think the +1 addition is a bug. For example, 
> consider the case where the extent in the y direction consists of a 
> single voxel situated at y = -1, and the Transform specifies that this 
> axis be reversed. This line should translate the origin to +1, so that 
> the voxel is newly situated at +1.
>
> In this case, the variables would be set as follows:
> original origin[1] = -1
> original spacing[1] = 1
>
> Transform =
> 1 0 0 0
> 0 -1 0 0
> 0 0 1 0
> 0 0 0 1
>
> yields
>
> transformedOrigin[1] = 1
> transformedSpacing[1] = -1
>
> transformedExtent[3] = 0
> transformedExtent[2] = 0
> i = 1
>
> The current code:
> origin[i] = transformedOrigin[i] + 
> transformedSpacing[i]*(transformedExtent[i*2+1] -  
> transformedExtent[i*2] + 1)
> = 1 + (-1 * (0 - 0 + 1)) = 1 - 1 = 0
>
> The old (correct?) code:
> origin[i] = transformedOrigin[i] + 
> transformedSpacing[i]*(transformedExtent[i*2+1] -  
> transformedExtent[i*2])
> = 1 + (-1 * (0 - 0)) = 1 + 0 = 1
>
> Is my understanding of this correct, or is there something I'm missing?
>
> Josh
>
>> Simon,
>>
>> I checked in your corrections this morning. I also fixed 
>> graphics/vtkVolume16Reader. Its logic was
>> also flawed.
>>
>> Thanks,
>>
>> Bill
>>
>> -----Original Message-----
>> From: Simon Warfield [mailto:warfield at bwh.harvard.edu 
>> <mailto:warfield%20at%20bwh.harvard.edu>]
>> Sent: Sunday, July 08, 2001 7:09 PM
>> To: vtkusers at public.kitware.com 
>> <mailto:vtkusers%20at%20public.kitware.com>
>> Subject: [vtkusers] bug in vtkImageReader.cxx ?
>>
>>
>>
>> I notice that if I read some image data using vtkImageReader and 
>> transform
>> it to correct for patient orientation using the 
>> vtkImagerReader::SetTransform
>> function, then the new origin is at an unexpected location.
>>
>> I get data looking like:
>> # vtk DataFile Version 3.0
>> vtk output
>> BINARY
>> DATASET STRUCTURED_POINTS
>> DIMENSIONS 256 256 90
>> SPACING 0.703125 0.703125 1.5
>> ORIGIN -90 -89.2969 -67.5
>> CELL_DATA 5787225
>> POINT_DATA 5898240
>>
>> When I expect to get data looking like:
>> # vtk DataFile Version 3.0
>> vtk output
>> BINARY
>> DATASET STRUCTURED_POINTS
>> DIMENSIONS 256 256 90
>> SPACING 0.703125 0.703125 1.5
>> ORIGIN -90 -90 -67.5
>> CELL_DATA 5787225
>> POINT_DATA 5898240
>>
>>  I think the problem is that line 936 of imaging/vtkImageReader.cxx 
>> should not
>> read
>>          (transformedExtent[i*2+1] -  transformedExtent[i*2]);
>> but instead should be:
>>          (transformedExtent[i*2+1] -  transformedExtent[i*2]+1);
>>
>>  This part of vtkImageReader.cxx in the latest vtk nightly is doing a 
>> shift of the origin when an axis is reflected, but seems to be using 
>> one less
>> than the width of the data to compute the new origin. I think the width
>> of the data should be used.
>>
>>  Could someone check this, and make the change to the CVS repository 
>> if it
>> should be fixed ?
>>
>> -- 
>> Simon Warfield, Ph.D. warfield at bwh.harvard.edu 
>> <mailto:warfield%20at%20bwh.harvard.edu> Phone:617-732-7090
>> http://www.spl.harvard.edu/~warfield 
>> <http://www.spl.harvard.edu/%7Ewarfield>           FAX:  617-582-6033
>> Thorn 329, Brigham and Women's Hospital, Harvard Medical School
>> Department of Radiology, 75 Francis St, Boston, MA, 02115
>>
>
> _______________________________________________
> This is the private VTK discussion list. Please keep messages 
> on-topic. Check the FAQ at: <http://public.kitware.com/cgi-bin/vtkfaq>
> Follow this link to subscribe/unsubscribe:
> http://www.vtk.org/mailman/listinfo/vtkusers