Cast Shadow and Reflection Image Suppressing of Dynamic Object Segmentation with Edge Features Kernel Density Estimation
Zheng Li, Hong Ma, Jian Bo Yang, Dongming Tang · 2007
In image analysis or vision understanding systems, the accuracy of segmentation results affects the quality of the systems seriously. The dynamic cast shadows and reflection images are ever-present fake objects in dynamic object segmentation, and they deteriorate the segmentation quality seriously. This paper presents a technique for cast shadow and reflection image suppressing of dynamic objects segmentation in videos. This technique is based on dynamic edge features and kernel density estimation. Unlike the classic kernel density estimation model which can only suppress cast shadows in color videos, this model can also suppress them in intensity videos, and with normal conditions it can suppress reflection images effectively. Although this technique introduces a little drawback of true object segmentation quality, its ability of fake objects suppressing is remarkable. Several experimental results with real videos are presented to demonstrate the effectiveness of this model.