Convex Active Contours with Locally Bias Field Estimator

Abstract

Intensity inhomogeneity or bias field in natural and medical images make image processing challenging. In this paper we have introduced a novel technique in which we first estimate the bias field using multi-scale filtering. Based on the bias field, the bias field corrected image is obtained which is used for accurate segmentation. For segmentation, a convex functional is proposed which is also suitable for multi-region segmentation. The proposed formulation is extended to vector-valued images and texture images. For comparison, the results are compared with state of the art models both qualitatively and quantitatively which validate strong enactment of the proposed formulation.