Piecewise-Constant Image Segmentation Model with Novel Data-Fitting terms
Miyoun Jung, Myeongmin Kang, Myungjoo Kang · 한국산업응용수학회 학술대회 논문집 · 2013
This article presents variational piecewise-constant image segmentation models that incorporate L norms as data-fitting measures. The L norms allow to segment images with low contrast or outliers such as impulsive noise. The regions to be segmented are represented as smooth functions restricted to [0, 1], instead of the Heaviside expression of level set functions. To handle both non-differentiable data-fitting and regularization terms, we first transform the proposed unconstrained minimization problems to equivalent constrained ones, and then apply a variant of the augmented Lagrangian method, the alternating direction method of multipliers, to solve the problems. This brings fast and efficient iterative algorithms for piecewise-constant image segmentation. The segmentation framework is extended to vector-valued images as well as to a multi-phase model to handle arbitrary number of regions. We show comparisons with ChanVese models that use the L fidelity terms.