Segmentation of objects from background with two stageautomatic thresholding

In this article, a new method is presented for automatic thresholding. We calledthis method two stage thresholding. In this technique, the concept ofreconstructing a different image is developed by segmenting the pixels of objectswhich have less distinctness against the background. In this article, this method isused for segmenting vehicles in a traffic scene by one approach of a signalizedintersection. The experimental results indicate the considerable improvement ofthe two stages of thresholding in comparison with the traditional thresholding.Intersection parameters such as queue length and volume can be measured basedon this procedure for intelligent traffic signal control.

Keywords:

two stage thresholding , reconstructed difference image , Tall Waterfall andShallow River model , within class variance , composition , median algorithm