背景重构是视频图像处理领域的支撑性工作之一.针对传统的背景重构算法运算复杂、背景图像失真等不足，本文提出了一种像素序列形态适应性背景重构算法.该算法通过提取像素序列形态特征进行分类处理，不同形态适用独立的背景提取策略、背景更新时刻和背景更新策略.实验结果验证表明：该算法无需对视频场景中的背景和运动目标建立模型，可直接从一组含有运动前景的视频图像中准确地重构背景，并有效避免混合现象；背景缓慢变化和突变时，亦可快速有效地完成背景重构.

Background reconstruction is a fundanmental task of video and image processing. To overcome disadvantages of traditonal backgound reconctruction algorithms, such as complex computation and distortional backgound, a pratical backgound reconstruction algorithm was proposed based on pixel sequence pattern classification. The pixel sequence patterns, extracted through calculating differences between sequential two frames, were classified by means of pixel sequence patterns’ charateristics determined by those differences. Furthermore, independent background extraction and background update mechanisms were destined for pixels with different pixel sequence patterns. Simulation results indicate that correct backgrounds of video with moving targets can be reconstructed without the models of background and moving targets. And, the background can be effiently updated when backgound is changed.