niftynet.layer.histogram_normalisation module

This class computes histogram based normalisation. A training process is first used to find an averaged histogram mapping from all training volumes. This layer maintains the mapping array, and the layer_op maps the intensity of new volumes to a normalised version. The histogram is computed from foreground if a definition is provided for foreground (by binary_masking_func or a mask matrix)

class HistogramNormalisationLayer(image_name, modalities, model_filename=None, binary_masking_func=None, norm_type='percentile', cutoff=(0.05, 0.95), name='hist_norm')[source]

Bases: niftynet.layer.base_layer.DataDependentLayer

__init__(image_name, modalities, model_filename=None, binary_masking_func=None, norm_type='percentile', cutoff=(0.05, 0.95), name='hist_norm')[source]
Parameters:
  • image_name
  • modalities
  • model_filename
  • binary_masking_func – set to None for global mapping
  • norm_type
  • cutoff
  • name
layer_op(image, mask=None)[source]
is_ready()[source]
train(image_list)[source]