vwshed − Watershed Image Segmentation

vwshed if=<ifile> of=<ofile> t=<threshold> [-v] [-b] [-2d] [-o]

vwshed applies the watershed transformation--a region-based segmentation technique--to segment a 2D or 3D image. The required input file is the image that needs to be segmented (if=). The output (of=) is the segmented image in type short (unless specified as byte by the -b flag). The threshold parameter (t=) specifies a "depth" that causes adjacent regions to be joined if both regions are "shallower" than t. See the -o flag below for an alternate option for t.

-2d is the flag which instructs vwshed to process a 3D image as a series of 2D images. -b is the flag which instructs vwshed to convert the output image to type byte. The -o flag provides an alternate option regarding the threshold value t, as it requires only one of two adjacent regions to be "shallower" than t for them to be joined. -v is the verbose flag.

vwshed is unique among watershed algorithms as it uses the traditional watershed method to segment the image but also utilizes a threshold to join regions based on their respective depths and reduce oversegmentation, the most notable problem with watershed image segmentation. The paper entitled, "An Efficient Watershed Algorithm Based on Connected Components" by Bieniek and Moga provides a step-by-step explanation of the fundamental watershed transformation and is published in Pattern Recognition 33:6, pp. 907-916, 2000.

vwshed is restricted to usage on 2D, 3D, 16-bit, and 8-bit images and only single-channel data. vwshed cannot process 16-bit images with a range larger than 32,767 between the maximum and minimum pixel values in the image. vwshed cannot process large images that have more than 32,767 regional minima among pixel values.

if=

input image

of=

output image

t=

threshold

-b

make output image as type byte

-2d

process image as series of 2D images

-o

combine adjacent regions if either is "shallower" than threshold t

-v

verbose mode

D. T. Masetti and A. P. Reeves