BinaryImageToShapeLabelMapFilter vs. LabelImageToShapeLabelMapFilter

Fijoy_Vadakkumpadan · August 20, 2018, 9:03pm

Hello,

What’s the best filter to use to compute size/shape attributes of all connected components in an image?

Looking at the docs, I see two filters, LabelImageToShapeLabelMapFilter and BinaryImageToShapeLabelMapFilter, for this purpose. But the latter seems to be a combination of the former and the ConnectedComponentImageFilter. If so, is there any other difference between the 2 filters? Specifically, do both filters give the user access to the same set of size/shape attributes of each connected component?

Thanks.

tim-evain · August 21, 2018, 10:08am

Hi Fijoy,

If I recall correctly the only difference is that the Binary one will treat all non-zero as the same “level”, i.e. if you have neighboring components of different non-zero values, they will be merged. Otherwise, ShapeLabelObjects are computed in both case, so the attributes will be strictly the same.

HTH,

blowekamp · August 21, 2018, 1:24pm

Hello Fijoy,

If you look a the source code for the two filters:

github.com

InsightSoftwareConsortium/ITK/blob/764472b182f764512549f6481a343c396ce28cee/Modules/Filtering/LabelMap/include/itkBinaryImageToShapeLabelMapFilter.hxx#L66


BinaryImageToShapeLabelMapFilter< TInputImage, TOutputImage >
::EnlargeOutputRequestedRegion(DataObject *)
{
this->GetOutput()
->SetRequestedRegion( this->GetOutput()->GetLargestPossibleRegion() );
}


template< typename TInputImage, typename TOutputImage >
void
BinaryImageToShapeLabelMapFilter< TInputImage, TOutputImage >
::GenerateData()
{
// Create a process accumulator for tracking the progress of this minipipeline
ProgressAccumulator::Pointer progress = ProgressAccumulator::New();


progress->SetMiniPipelineFilter(this);


// Allocate the output
this->AllocateOutputs();


typename LabelizerType::Pointer labelizer = LabelizerType::New();

github.com

InsightSoftwareConsortium/ITK/blob/764472b182f764512549f6481a343c396ce28cee/Modules/Filtering/LabelMap/include/itkLabelImageToShapeLabelMapFilter.hxx#L64


LabelImageToShapeLabelMapFilter< TInputImage, TOutputImage >
::EnlargeOutputRequestedRegion(DataObject *)
{
this->GetOutput()
->SetRequestedRegion( this->GetOutput()->GetLargestPossibleRegion() );
}


template< typename TInputImage, typename TOutputImage >
void
LabelImageToShapeLabelMapFilter< TInputImage, TOutputImage >
::GenerateData()
{
// Create a process accumulator for tracking the progress of this minipipeline
ProgressAccumulator::Pointer progress = ProgressAccumulator::New();


progress->SetMiniPipelineFilter(this);


// Allocate the output
this->AllocateOutputs();


typename LabelizerType::Pointer labelizer = LabelizerType::New();

You will see they are both composite filters. Which means the filter is just composed of other filters connected. The BinaryImageToShapLabelMapFilter is composed of the BinaryImageToLabelMapFilter->ShapeLabelMapFilter, while the LabelImageToShapeLabelMapFilter is make of the LabelImageToLabelMapFilter->ShapeLabelMapFilter. The second filter is the one which computes the statistics on a LabelMap, so that step will be the same between the two filters you mention.