Coefficient estimation of the energy functional area term

Many applications of computer vision, such as variation frameworks, operate using level set methods which require some unknown parameters to be chosen before evolution of the level set function. In general, the parameters should be estimated using provided data, however, in some cases, it can be defined empirically. The present work focuses on estimation of the coefficient in the energy functional that computes a weighted area of the region inside the contour and speeds up its motion toward the object boundaries. The paper discusses a new approach for the coefficient estimation comprised the image features such as mean and variance values of pixel intensities and image gradients. The advantages of the precise estimation of this parameter are following: (1) the convergence of the evaluation process is getting faster if the value of the coefficient in the weighted area term is higher that, therefore, may speed up the curve evolution; (2) the contour may pass through the object boundary in some lower contrast images if the coefficient is too large, thus the calculation of the coefficient may avoid this effect called boundary leakage. The provided result shows that the suggested approach on parameter estimation can increase the speed and quality of the convergence driving the motion of the zero level curve in images with different contrast.