A shape-initialized and intensity-adaptive level set method for auroral oval segmentation

Abstract

Auroral oval segmentation is of great significance in the field of spatial physics. However, random noise, low contrast and dayglow contamination in ultraviolet imager (UVI) images lead to various difficulties and challenges. To address these issues, in this paper we develop a shape-initialized and intensity-adaptive level set method. Firstly, morphological component analysis (MCA) based preprocessing is applied to reduce noise effect on UVI images. Subsequently, a saliency morphological map is constructed to capture the shape knowledge for auroral oval and used as the initial curve for level set. Finally, an intensity-adaptive level set evolution is implemented to construct the auroral oval boundary. Experimental results demonstrate that the proposed method is superior to the existing auroral oval segmentation methods for both full oval and gap oval images. In addition, the subjective and objective evaluations for segmentation results further validate the favorable performance of the proposed method.