A new motion segmentation method for dynamic scenes

作者：Xie Zhenping^*; Wang Shitong

来源：Information Technology Journal, 2011, 10(10): 1942-1949.

DOI：10.3923/itj.2011.1942.1949

• Common knowledge
• Dynamic scenes
• Gaussian Mixture Model
• Image features
• Level Set
• Level Set method
• Motion segmentation
• Motion tracking
• Moving objects
• Segmentation boundaries
• Segmentation methods
• Statistical properties

Because of its requirement of precisely extracting moving objects, motion segmentation especially for dynamic scenes is more difficult than motion tracking. So, efficient image segmentation methods may be employed to solve above problem, which drives us to develop more novel motion segmentation method for dynamic scenes. In this study, a novel image segmentation method using level set is employed to design a new motion segmentation method for dynamic scenes. From theoretical analysis, level set method and Gaussian Mixture Model (GMM) are two very valuable tools for natural image segmentation. The former aims to acquire good geometrical continuity of segmentation boundaries, while the latter focuses on analyzing statistical properties of image feature data. Derived from this common knowledge, a novel level set image segmentation method integrated with GMM (called as GMMLS) has been proposed in previous studies. Wherein, Gaussian mixture model is used to analyze image feature, moreover the effectiveness and good performance of GMMLS also have been demonstrated. Based on GMMLS, a new motion segmentation method for dynamic scenes is proposed in this study and experimental results on several moving objects in dynamic scenes indicate that new method owns some excellent and particular worthiness on such practical applications.

• 出版日期2011
• 单位江南大学