Monitored video background image modelling method

The invention provides a monitored video background image modelling method. An excellent subjective quality can be guaranteed by adopting the boundary detection method of block and the gradient calculation of image residuals and the length adjustment method of a self-adaption training set. The background image modeling method is based on a block, and YUV components are stored, the generated background images can guarantee the high-quality chromacity and brightness consistency. The modelled background images are used as global reference images for video coding, so that the bite rate can be obviously saved, and the background image for reference of global coding can be established while the subjective quality is guaranteed. 本发明提供种监控视频背景图像建模方法，采用块的边界检测法和图像残差的梯度计算以及自适应训练集的长度调节法，保证了优异的主观质量。本发明的背景图像建模方法是基于块的，同时存储了YUV三分量，生成的背景图像保证了高质量的色度和亮度的致性。将建模后的背景图像作为全局参考图像进行视频编码，可以明显地节省比特率，在保证主观质量的同时建立可供全局编码参考的背景图像。