Abstract: Based on the iterative graph cut method, this paper uses image pyramid to accelerate the speed of segmentation and introduces a balance factor to determine the significance between the color and depth information. This method can segment the depth image easily and precisely with the pyramid and the balance factor. The balance factor is the measure of the discrimination for the color and depth information. If the depth information is sufficient to distinguish the object from the background, then the depth information plays a more important role in the segmentation process, otherwise the color information takes more important role. By using the image pyramid our method can interact with the iterative graph cuts and greatly reduces the processing time without influencing the segmentation precision.

Key words: iterative graph cuts, balance factor, Gaussian mixture, depth histogram

摘要： 以深度图像为分割对象，在迭代图割算法的基础上，通过引入分层机制加快图割执行速度，并通过引入平衡因子来平衡颜色纹理和深度之间的重要程度，从而有效地对深度图像进行分割。利用平衡因子可以在深度信息能够明显区分前背景的情况下，重点利用深度信息来分割图像，反之则重点利用颜色和纹理信息。而在迭代图割算法中，分层机制的引入能够在不降低分割精度的情况下有效地减少图割算法的执行时间。

关键词: 迭代图割, 平衡因子, 混合高斯, 深度直方图