Delineating trees in noisy 2D images and 3D image-stacks

We present a novel approach to fully automated delineation of tree structures in noisy 2D images and 3D image stacks

German Gonzalez


Scholarcy highlights

  • Tree-like structures appear at many different scales and in many different contexts
  • We argue that this is in large part because current approaches are far too local and do not sufficiently take into account the global tree structure when making decisions
  • Since the ground truth contains width estimates, we render an image stack by assigning a one to all points that are within the corresponding distance from the dendritic spine and zero to the others
  • While spanning all anchor points yields many false positives highlighted by the yellow ellipses in Fig. 7(c) which are hard to prune at a later post processing step, our K-MST based approach eliminates many outliers through its global optimization procedure. This observation is supported objectively by the ROC curves given in Fig. 4(b),(c), where MST yields a false positive rate almost twice as high as that of our algorithm with an increase of only 7% in the true positive rate
  • We have presented a method to automatically infer tree structures present on 2D and 3D images
  • They fit perfectly in our framework and should further boost performance

Need more features? Save interactive summary cards to your Scholarcy Library.