【作者】 李剑峰；

【导师】 乐光新；

【作者基本信息】 北京邮电大学 ， 通信与信息系统， 2008， 博士

【摘要】 21世纪人类进入信息社会,知识和信息爆发式产生,人们已不满足于仅通过报刊、电话通信等来获取文字、语音等信息,更希望能够通过图像通信,尤其是通过无线数字图像通信来随时随地了解和获取一些实时的现场视频。因此,高效率、高清晰度、高可靠性的数字视频图像处理与传输系统的研究,尤其是在无线移动环境中的研究,具有重要的理论和现实意义。基于此,论文主要研究无线数字图像通信系统中的高速率无线传输和增强处理两大部分。数字图像的无线传输:为了更好地对抗无线多径衰落信道的影响,论文给出了一种基于OFDM技术的高速无线数字图像传输系统的设计方案,并研究了基于功率谱估计的信道估计算法。该估计算法得到的信道估值较最小均方误差算法更准确,系统性能更接近于接收端已准确获知信道传输系数时的性能。数字图像增强处理:首先,提出一种基于形态滤波的数字图像增强算法。该技术有机结合神经网络结构和形态滤波理论,推导出一种利用神经网络实现灰度形态滤波器参数的优化设计方法,有效解决了形态滤波器结构元素难以自适应跟踪复杂变化的图像模型的问题。该方法显著提高了形态滤波的性能,且设计简便,实用性强。然后,基于信息融合理论,提出一种改进型的D-S证据理论的组合准则,并给出一种基于改进型的D-S证据理论的决策层融合数字图像滤波算法。该算法主要针对数字图像的线性、非线性混合滤波算法中存在的边界点单源判断误报风险大、可靠性和容错性差的缺陷,将边界点的判决由单一准则变为多个准则,再运用智能化、容错能力强的多源数据信息融合技术,提高了边界点判断的准确性。且原始图像越差,滤波性能的提升就越大。尤其是当边界区域模糊或图像质量不好时提升效果更佳。最后,结合离散傅立叶变换的Cooley-Tukey迭代算法和修剪算法,提出一种针对频域图像处理的改进的FFT并行迭代算法,并对算法复杂度、平均时延及其所引起的时延抖动等进行估算,给出近似估算公式。与传统FFT快速算法和Cooley-Tukey迭代算法相比较,该算法复杂度低、运行速度快,且易于硬件实现。更多还原

【Abstract】 With the arrival of 21 century, which is also known as information society century, knowledge and information generate explosively. People have not been satisfied with obtaining information of text, speech, etc, via the media of magazine and telephone, and hope to get real-time field video information anywhere and anytime through digital image communication. Therefore, the research on digital video processing and transmission systems with high-efficiency, high-definition, and high-reliable is significantly important. So, the research works of the thesis mainly focus on high-data-rate wireless transmission and enhancement processing in wireless digital image communication systems.Wireless Transmission of Digital Video Image:In order to overcome the effects of multi-path fading over wireless channels, a wireless high-data-rate digital image communication system based on orthogonal frequency division multiplexing modulation technique is proposed, and a multi-path channel estimation algorithm based on power-spectrum-density (PSD) estimation is presented. Simulation results show that the mean-square-error between the estimated channel by PSD and the known channel is less than that of the least square estimation and the minimum-mean-square-error (MMSE) estimation, and the performance of the OFDM system approaches to that of the system with ideal channel information.Enhancement processing of Digital Image:First, a digital video enhancement algorithm based on morphological filter is proposed, which combines the neural network structure and morphological filtering theory. An optimization method is derived to implement the gray morphological filter, which resolve the problem that the structure of morphological filter is hard to track adaptively the complicated variation of video model. This method improves the performance of morphological filter, and is easy to implement.Then, an improved combinatorial criterion of advanced D-S theory of evidence based on information fusion is proposed, and a decision-level fusion filtering algorithm for digital image is presented, which resolves the problems of high-risk, low reliability and low-error tolerance in the decision based on the boundary data of single-source. It improves the precision of the decision based on the boundary data by changing the decision rules for the boundary data from single criterion to multi-criterions and using intelligent multi-source data fusion technique, which has high error-tolerance and improves the decision precision. The results show that the worse of the original image, and the better the improvement of the filtering performance. Especially when the boundary region is blurred or the quality of the image is bad, the performance improvement is more obvious.Finally, combined with the Cooley-Tukey iterative and cut algorithm of discrete Fourier transform, an improved parallel FFT iterative algorithm for image processing in frequency domain is presented. The complexity of the algorithm, average delay and the induced delay jitter are evaluated and an approximate evaluation formula is also derived. Compared with traditional FFT fast algorithms and the Cooley-Tukey iterative algorithm, this algorithm has low complexity, fast running time, and low complexity to implement.更多还原