【作者】 黄维；

【导师】 常义林；

【作者基本信息】 西安电子科技大学 ， 通信与信息系统， 2009， 硕士

【摘要】 为了获得良好的视频质量和较高的压缩率,视频预处理技术在视频处理与通信应用中占据重要的地位。为了平滑视频中同一镜头图像的亮度,提高视频编码效率,论文深入研究了从视频采集到采集输出之间的背景闪烁消除技术;为了提高图像空间分辨率,论文对图像插值技术研究做了详细的讨论。在两项技术研究中,分别提出了新的解决方法。背景闪烁是由于光线环境和相机的自动曝光引起场景图像的明暗变化,该现象严重影响了视频的观看效果和压缩效率。本文提出一种新的消除背景闪烁的方法,算法寻找视频序列中的亮度参考帧,使用最小二乘法确定出当前帧和参考帧之间的背景亮度响应函数,使用亮度补偿的方法消除背景闪烁。实验结果表明该算法可以有效地消除背景闪烁现象,用于编码传输可以节省20%-30%的比特率。研究成果已经应用到华为公司的多媒体产品中。针对传统图像插值算法存在的模糊和锯齿问题,本文提出一种新的基于方向判断的加权图像插值方法。该方法计算四个不同方向的梯度,根据梯度值确定插值方向和每个方向的加权系数,使用加权插值得到最终的图像值。实验结果表明,该方法可以获得较高质量的高分辨率图像。更多还原

【Abstract】 In order to obtain high video quality and high compression performance, video preprocessing technique plays more important role in video processing and communications. In this dissertation, two processing modules between the video sensor and the output of the capture device are deeply studied. We pay more attention on the background flicker reduction and image interpolation.Background flicker which is caused by varied light environment and the auto exposure function of video cameras can extremely degrade the video quality and compression efficiency. A new luminance flicker reduction algorithm is proposed in this paper under the circumstances. Firstly, the reference picture for illumination compensation is found out in the proposed algorithm, and then the parameters of the response function of background luminance between the reference picture and the current picture is obtained using the least-square method. Finally, the luminance flicker can be reduced by illumination compensation. Experimental results show that the background luminance flicker can be effectively reduced using this algorithm and bit rate can be saved by 20%-30% when the compensated video sequences are encoded and transmitted. The developed algorithm has been successfully applied to video products of Huawei limited technology cooperation company.To remove blurring effect and zigzag edges occurring in traditional interpolated images, a new algorithm based on direct-judgment is proposed in his paper. The basic idea is to compute the gradient values of four directions to determine the interpolated directions and its corresponding weighted coefficients, then, the final image is obtained using weighting interpolation. The experimental results show that relatively higher quality images with high definition are achieved.更多还原