SplineSnake

An ImageJ Plugin for Image Segmentation using Closed Snake (Active Contour)

Author: Mathews Jacob

Description

SplineSnake is an implementation of the snake algorithm for the segmentation of graylevel images. The snake is a spline curve that is attracted by image information to the object boundaries. The formulation ensures that the extracted boundary is smooth. The implementation is based on [1].

Download

This distribution is dated March 27, 2003. It includes the precompiled class files. Please contact me if you require the source files.

To install, extract the "SplineSnake" folder to the "Plugins" directory of ImageJ. You need to restart ImageJ for the plugin to be active. Please drop me a line if you are successfully using SplineSnake.

Using the Plugin

Open the image using ImageJ and launch the plugin. To work on a new image, one has to exit SplineSnake and restart the plugin.

		Add points: Click on the image to start the curve input. Double click or click on the first(red point) to close the curve
		Delete points: Click on the knot to delete it.
		Delete all: Clear everything.
		Add constraints: Click on the point through which the curve should pass through.
		Delete constraint: Click on the constraint point to delete it. Double click the button to clear all constraints
		Preferences: See below for explanations of the parameters.
		Snake: Start the snake algorithm.
		Accept: Accept the final snake (red) as the initialization.
		Stack: Reserved for future use on image stacks
		Save to File: Saves the current snake to a file.
		Exit SplineSnake: Exit the SplineSnake to ImageJ
		Zoom.

The snake parameters and their explanations are given below. The important terms (the ones you may want to play with) are marked in red.

The parameters on the left are the ones used during the curve initialization. They are

Optimization: Level of optimization during curve initialization.
Knot spacing: Spacing of the curve knots. Widely spaced knots imply smooth curves, but may fail to capture the details. Its proper choice depends on the application.
StepSize: Size of the step in the optimization during initialization.
Smoothing radius: used for smoothing the image prior to the gradient computation.

The ones on the right are used by the snake.

Image energy: proper linear combination of gradient and region energies can result in better convergence. The right combination depends on the image.
Max #iterations: Maximum number of iterations.
Accuracy: Accuracy to which the snake is optimized.
Smoothing radius: used for smoothing the image prior to the gradient computation.
Spring weight: Weight for the constraint (active only if Spring Constraint is set. A high value may cause the snake to misbehave).

The current parameter values may be saved to avoid entering them each time. The saved values can be recovered by pressing the "Default" button.

References

[1] M. Jacob, T. Blu, M. Unser, "A Unifying Approach and Interface for Spline-Based Snakes " , Proceedings of the SPIE International Symposium on Medical Imaging: Image Processing, San Diego CA, USA, February 17-22, 2001, vol. 4322, Part I, pp. 340-347.

[2] M. Jacob, T. Blu, M. Unser , "Efficient energies and algorithms for parametric snakes", Submitted to IEEE Transactions on Image Processing.

Back	Last updated on March 27, 2003.