数字多媒体视频目标区的智能拾取仿真研究

Research on Intelligent Picking-up Simulation of Digital Multimedia Video Target Area

数字多媒体视频作为一种采集输入视频流,具有严重的动态波动与噪声干扰等特点,现有方法很难对其中的目标区进行精确快速的拾取。为解决上述问题,提出了融合像素空间关系与阴影判别的混合高斯模型。为了尽可能完整的描述多媒体视频帧,采用帧图像的像素特征建立混合高斯模型,来描述帧图像的信息,并根据更新计算,实时得到帧图像的动态变化,通过像素状态从而剥离出目标区。由于视频流易受噪声影响,根据像素空间关系得出邻域像素间的约束条件,从而避免噪声像素带来的干扰,另外针对目标区拾取易受物体阴影影响,采用HSV三要素过滤帧图像中的阴影像素。通过仿真结果,证明了融合像素空间关系与阴影判别的混合高斯模型能够有效应对数字多媒体视频的复杂特性,适应性更好,且具有良好的拾取精度与速度。

As an Input Video Stream, digital Multimedia Video has the characteristics of serious dynamic fluctuation and noise disturbance. It is difficult for existing methods to pick up the target area accurately and quickly. For this reason, a Mixed Gauss Model(GMM) is proposed, which combines the spatial relationship of pixels and shadow discrimination. In order to describe multimedia video frames as completely as possible, the GMM was established using the pixel features of the frame image to describe the information of the frame image. According to the update calculation, the dynamic change of frame image can be obtained in real time. The target area was stripped out through the pixel state. Because video streams are susceptible to noise, according to the spatial relationship of pixels, the constraints among neighboring pixels were obtained in order to avoid the interference caused by noise pixels. In addition, target area pickup is vulnerable to object shadows, HSV three elements were used to filter negative pixels in frame images. Through the simulation results, it was proved that the Mixture Gauss Model, which combines the spatial relationship of pixels and shadow discrimination, can effectively deal with the complex characteristics of digital multimedia video and has better adaptability. And it has good picking accuracy and speed.