Gradient Vector Flow

The Gradient Vector Flow tool allows calculating the image gradient in a specified direction [1], which is for instance used for active contours algorithms.

Different algorithms are supported:

• GVF: calculates the gradient using the gradient vector flow algorithm;
• General GVF: calculates the gradient using a generalized gradient vector flow algorithm;
• Differential gradient: calculates the gradient by subtracting values of neighbouring pixels;
• Sobel operator: calculates the gradient by convolving filter mask with matrix consisting of image pixels.

During the estimation procedure the following parameters can be set:

• Gradient width: distance (in pixels) between two points, the intensity difference which defines the gradient value;
• Iterations: number of iterations performed during GVF and GGVF calculation;
• Smoothing parameter (m): regularization parameter governing the trade-off between the first and the second integral term. Smoothing parameters should be set according to the amount of noise present in the image: the higher the noise the bigger the value.
• Time step (D): is representing the time length for each iteration.

In order to guarantee algorithm convergence, the smoothing parameter and the time step should satisfy the following expression: m<-1.36*Dt+0.22. Therefore, the Parameters restriction box should be enabled.

By enabling Image Preview and the Show gradient field, the gradient vectors can be visualized.

Reference

[1] Chenyang Xu, Jerry L. Prince. Snakes, Shapes and Gradient Vector Flow, Transactions on Image Processing, March 1998, p. 359-369.