[Insight-users] Filters on 3D images

[Luis Ibanez]

Hi Ramakrishna



 From your code it looks like your pipeline is:


     Read ->

     GaussianFilter -> Threshold ->

     ZeroCrossing -> Laplacian -> Write



Is this right ?


--


Some comments about it:

1) the RecursiveGaussianImageFilter may not be
    the filter you want in here.  This filter
    computes the equivalent of a convolution
    with a gaussian kernel but does this using
    IIR filters. The filter only applies the
    processing in *one* of the N-Dimensions.
    This filter is not intended to be used alone
    but rather in tandem to apply different operations
    along different dimension.

    If what you want is to blurr the image with
    a gaussian kernel, you may want to use the
    DiscreteGaussianImageFilter:
 
http://public.kitware.com/Insight/Doxygen/html/classitk_1_1DiscreteGaussianImageFilter.html 


    which performs a traditional convolution taking
    advantage of the separability of the Gaussian.


2) By thresholding the gaussian you are creating
    artificial contours on the image. If your purpose
    on the pipeline is to detect contours, this
    thresholding here may not be helping much.
    I would suggest you to remove it from the pipeline.


3) You are applying the ZeroCrossing filtering
    before the Laplacian. The output of Zero crossing
    is a binary image.  It is usual to apply the
    Laplacian first and then take the Zero crossings
    of the laplacian as contours. Probably you may
    want to revert the order of this two filters.



So,
the suggestion is to try the following pipeline:


Read->DiscreteGaussian->Laplacian->ZeroCrossing->Write


Please let us know if you have further questions,


   Thanks


    Luis


========================================


cspl wrote:

> Hi All,
> 
>  
> 
> I am working on Laplacian filter.I used this filter for two dimension 
> images ,it is working fine.But applying on multiple slices in an 
> image(.dcm) filters are not effective on all images.
> 
> I have writen code as follows.Please give me suggestion.
> 
>  
> 
> LPSTR __stdcall Laplacian(LPSTR FileName,long Thrval)
> {
>  GblCount++;
>  DICOMReader TempReader;
>  TempReader.StaticVarInit();
>  bool Flag = TempReader.ReadDicomImage(FileName);
> 
>  
> 
>  char *rep=TempReader.GetDicomInfo(cs_pixel_representation);
>  char *bits=TempReader.GetDicomInfo(cs_bits_allocated);
>  char *acquistion=TempReader.GetDicomInfo(cs_modality);
>  int represent=atoi(rep);
>  int type=atoi(bits);
> 
>  
> 
>  unsigned short * Buffer = (unsigned short *) 
> TempReader.GetAllImageData(); 
> 
>  
> 
> 
>  typedef itk::Image<unsigned short,3> ImageType;
>  ImageType::Pointer vol_init = ImageType::New();
> 
>  
> 
> 
>  ImageType::SizeType sz;
>  sz[0] = TempReader.width;
>  sz[1] = TempReader.height;
>  sz[2] = TempReader.Slices;
> 
>  
> 
>  
> 
>  
> 
>  ImageType::IndexType idx;
>  idx[0] = 0;
>  idx[1] = 0;
>  idx[2] = 0;
> 
>  
> 
>  
> 
>  
> 
>  ImageType::RegionType reg; 
>  reg.SetSize(sz);
>  reg.SetIndex(idx);
> 
>  
> 
>  vol_init->SetRegions(reg);
> 
>  
> 
>  vol_init->Allocate();
> 
>  
> 
>  
> 
>  
> 
>  for(unsigned int z=0; z<TempReader.Slices; ++z )
>  {
>   idx[2]=z;
> 
>  
> 
>   for (unsigned int y = 0; y < TempReader.height; ++y)
>   {
>    idx[1] = y;
>    for (unsigned int x = 0; x < TempReader.width; ++x)
>    {
>     idx[0] = x;
>     vol_init->SetPixel(idx, *(Buffer 
> +(z*TempReader.height*TempReader.width)+y *TempReader.width +x) );
>    }
>   }
>  }
> 
>  
> 
>  
> 
>  
> 
>  
>  typedef unsigned short PixelType;
>  typedef itk::Image<PixelType, 3> myImage;
> 
>  
> 
> 
>  itk::RecursiveGaussianImageFilter<myImage, myImage>::Pointer
>  gaussianFilter = itk::RecursiveGaussianImageFilter<myImage, 
> myImage>::New();
>  gaussianFilter->SetInput(vol_init);
> 
>  
> 
>  ///Laplician Filter
>  itk::ThresholdImageFilter< myImage >:: Pointer 
> thr=itk::ThresholdImageFilter<myImage>::New();
>  thr->SetInput(gaussianFilter->GetOutput());
>  if(strcmp(acquistion,"MR")==0)
>  {
> 
>  
> 
>   if(type==8 && (represent==1 ||represent==0))
>   {
> 
>  
> 
>    AfxMessageBox("type 16"); 
>    thr->ThresholdOutside(0,(unsigned short)(Thrval)) ;
> 
>  
> 
>   }
>   if(type==16 && (represent==1 || represent==0))
>   {
>    AfxMessageBox("type 16");
>    thr->ThresholdOutside(0,(unsigned short)((Thrval))*100 ) ;
>   }
>   else
>   {
>    AfxMessageBox("else");
>    thr->ThresholdOutside(0,(unsigned short)(Thrval)) ;
>   }
> 
>  
> 
>  }
> 
>  
> 
>  if(strcmp(acquistion,"CT")==0)
>  {
> 
>  
> 
>   if(type==8 && (represent==1 ||represent==0))
>   {
>    
>    thr->ThresholdOutside(0,(unsigned short)(Thrval)) ;
> 
>  
> 
>   }
>   if(type==16 && (represent==1 || represent==0))
>   {
>    
>    thr->ThresholdOutside(0,(unsigned short)((Thrval))*250 ) ;
>   }
>   else
>   {
>    
>    thr->ThresholdOutside(0,(unsigned short)(Thrval)) ;
>   }
> 
>  
> 
>  }
> 
>  
> 
> 
>  itk::ZeroCrossingImageFilter<myImage, myImage>::Pointer
>  zeroFilter = itk::ZeroCrossingImageFilter<myImage, myImage>::New(); 
>  zeroFilter->SetInput(thr->GetOutput());
> 
>  
> 
>  zeroFilter->Update();
> 
>  
> 
> 
>  itk::LaplacianImageFilter<myImage, myImage>::Pointer
>  lapFilter = itk::LaplacianImageFilter<myImage, myImage>::New();
>  lapFilter->SetInput(zeroFilter->GetOutput());
> 
>  
> 
>  lapFilter->Update();
> 
>  
> 
>  myImage::Pointer  imageo = lapFilter->GetOutput();
>  myImage::IndexType idx1; 
> 
>  
> 
>  for(unsigned int z=0; z<TempReader.Slices;++z)
>  {
>   idx1[2]=z;
>   for (unsigned int y = 0; y < TempReader.height; ++y)
>   {
>    idx1[1] = y;
>    for (unsigned int x = 0; x < TempReader.width; ++x)
>    {
>     idx1[0] = x;
>     *(Buffer +(z * TempReader.height *TempReader.width) +y 
> *TempReader.width +x)=imageo->GetPixel(idx1);
> 
>  
> 
> 
>    }
>   }
>  }
> 
>  
> 
> 
>  memcpy(TempReader.AllBuffers,Buffer,TempReader.width*TempReader.height*TempReader.Slices*sizeof(unsigned  
> short));
> 
>  
> 
>  LPSTR Temp = TempReader.WriteDicomFile();
>  int Len = strlen(Temp);
> 
>  
> 
>  LPSTR FName = new char[Len + 1];
>  strcpy(FName ,Temp);
>  return FName;
> }
> 
> Regards,
> 
> Ramakrishna
>