【作者】 周津；

【导师】 姚素英；

【作者基本信息】 天津大学 ， 微电子学与固体电子学， 2008， 博士

【摘要】 随着计算机技术、通讯技术以及超大规模集成电路技术的进步和相互融合,电视行业正在经历一场由模拟电视全面向数字电视转换的发展过程。由于数字电视巨大的市场潜力,作为数字电视系统核心的数字视频图像后处理芯片及相关技术的研发,成为了行业关注和竞争的焦点。对于高清晰度数字电视来说,视频后处理技术中最为重要的部分就是视频信号的去隔行处理功能,但已有的去隔行算法已经无法适应高清视频图像处理高性能、低成本的要求,因此本文针对高清数字电视应用进行了去隔行处理关键技术的深入研究。本文是天津市重大科技攻关计划项目“视频信号处理芯片的研发”课题的延伸,并与晶宝利公司合作进行高清数字电视后处理芯片的产品化开发。本文提出了结合电影模式处理和图像预降噪的高清视频去隔行处理系统,其主要创新点为去隔行插值、电影模式处理和隔行图像预降噪等三方面关键技术的新型算法研究。本文提出了基于全局统计和边缘平滑滤波的新型视频去隔行图像插值技术。该技术采用全局统计运动检测技术获取运动信息,同时结合边缘平滑滤波的场内相关性插值技术,进行高清视频图像自适应插值计算。实验结果表明,该技术的算法复杂度适当,对动态和静止图像都有较好的处理效果。文章提出了一种新型全格式电影模式处理技术。该技术利用逐像素的帧、场变化统计以及基于高斯分类的电影模式状态识别对电影模式源进行检测和处理。在实现方法上采用软硬件协同处理的方式,节省硬件资源占用的同时保证了处理模式的可更新性。实验结果表明,该技术可以有效识别各种电影模式序列以及影视混合图像并进行处理,图像还原结果得到有效保证。本文还提出了隔行图像自适应预降噪技术。该技术根据去隔行插值处理的数据结构,利用噪声的概率分布特征进行统计和分析,有针对性的采用场内滤波和场间滤波相结合的自适应降噪处理方法,有效的降低了常见高斯噪声和椒盐噪声的影响,提升了图像质量。最后,论文给出了高清视频去隔行处理系统的芯片架构和功能模块设计,并且定义了模块间的时序关系。仿真测试结果显示本系统设计满足高清视频处理时序要求,整体系统资源占用适当,适合于高清数字电视视频后处理芯片的实时应用。更多还原

【Abstract】 As the rapidly growth and combination of computer science, communicating and VLSI technology, TV industry is experiencing the thorough change from analog TV broadcasting system to DTV system. The research and development of digital video image post-processing IC and algorithm, which is the kernel of DTV system, became the main competition field of the huge future market. The most important function of high definition (HD) TV post-processing is the de-interlace processing of interlaced signal. Because the high performance and low cost of HD video can’t be acquired by current de-interlacing method, this dissertation mainly focuses on the deeply research of de-interlacing key technique for HDTV application.This thesis is an expansion of the project“Research and Development of Video Signal Processing Chip”which is supported by Tianjin key technologies R&D Program and the product development is cooperated with PowerLayer Microsystems (Beijing) Ltd. This dissertation introduces HD video de-interlace processing system with film mode processing and image pre-denoising. The creative points of this dissertation are three key techniques, included: de-interlacing interpolation, film mode processing and image pre-denoising.A novel HD video image de-interlacing interpolation technique which is based on global statistic processing and edge smooth filtering is proposed. In this technique, the exactly motion information is calculated by motion detection with global statistic. The relativity directional filtering with edge smooth is adopted by intra-field interpolation. The compute complexity of this HD video adaptive technique is advisable and the high quality results are achieved from both of motive and stationary object.New technique of film mode processing for all film cadences is proposed. The pixel level frame/field difference statistic is used by film mode detection based on gaussian classifier. The design of cooperation with software calculation and hardware operation saves the hardware resource and provides the update ability of the unknown cadences. As the result, all the existing film cadences and Video-on-Film (VoF) cases are correctly detected and processed.An interlaced signal adaptive pre-denoising technique is proposed. In this technique, the noise probability statistical characteristic is acquired and analyzed in terms of the data structure of interlaced image. The pixel is selective processed by intra-field filtering and inter-field filtering. This adaptive processing effectively reduced the Gaussian noise and Salt-Pepper noise and improved the video image quality.At last this dissertation provides the VLSI architecture of HD video de-interlace processing system and hardware design of functional module and interfaces. The simulation reports show that the HD video timing requirement is satisfied and hardware resource usage of the whole system is reasonable. This system is suitable for real-time application of HDTV post-processing IC.更多还原