2014 Summer Project Week:Multiscale Non Local Means filter (NLM) for chest CT images

From NAMIC Wiki
Jump to: navigation, search
Home < 2014 Summer Project Week:Multiscale Non Local Means filter (NLM) for chest CT images

Key Investigators

  • Pietro Nardelli
  • Raul San Jose Estepar

Project Description


  • Improving the quality of CT images is important due to the increased use of these images for diagnostic purposes and the need to reduce dose. We have inspected possible methods for improving image quality while preserving important information. We are thus interested in developing a modified NLM formulation based on quadratic approximation of distances. Our idea is to exploit a multi-scale approach in order to better compute salient features that will be used in the approximation of local patches centered on pixels under comparison. The quadratic approximation of local intensities is very suitable in modalities like CT where tubular and sheet-like structures are visible. Our interest is focused mainly on chest CT images, as the filter would be used combined with an airway segmentation module. Our goal is thus to develop a NLM filter to be integrated within Slicer. The filter should also be tested on different noisy images.

Approach, Plan

  • A first version of a modified NLM filter has been already developed as a standalone module using ITK classes. However, it is still not working properly. Therefore, the module needs to be improved first in order to obtain good results.
  • The module will be then integrated within Slicer as CLI/Scripted module.
  • A new Slicer interface will be created in order to make the module more user-friendly


  • Small issues have been fixed.
  • The filter has been tested on a few chest CT images
  • More improvements and new tests will be necessary.