Pulse Coupled Neural Network Segmentation Algorithm

Biomedical Optics Laboratory Segmentation Code Directory Contents:

Calibration Data

True-positive and false-positive sets for neural network training. Kept for records. If applied to a new set of images, it is best to use your own training images for best results.

Cropped Images for Hand Segmentation

Hand Segmented Images, used for comparison as a standard.

GMRF

Gaussian Markov Random Field Segmentation Algorithm (Luck et al. 2005)

HS_Script

Outputs segmentation analysis for given image. Use HS_script.

Image Model

Image model generator. Use imodel.m to generate initial model, this creates a *.mat file that is used by imodel2.m. Use imodel2 to add contrast and noise, you must have a *.mat file generated or the script will fail. The image model generates a "8-bit" image in a 16-bit container, so output may not be visible to the user (This allows for pixel values greater than 255).

Overlap

Compares segmentation results between two images. Defines object/pixel overlap. Use overlap.m

SCM_GUI_V2_FINAL

SCM segmentation algorithm. Use SCM_seg.m

Training Images

Training images at 4 depths.