【作者】 张北斗；

【导师】 马义德；

【作者基本信息】 兰州大学 ， 电路与系统， 2007， 硕士

【摘要】 现代图像处理理论研究表明，新时期的数字图像处理技术要向高速度、高质量、智能化方向发展，且能够模拟生物视觉系统的处理过程。“第三代神经网络”—脉冲耦合神经网络(Pulse Coupled Neural Network，PCNN)来源于哺乳动物猫的视觉皮层神经细胞的研究成果，通过对其同步脉冲发放现象进行研究，产生了PCNN神经元模型。由于PCNN是对哺乳动物视觉系统较为精细的模拟，因此它更接近视觉系统处理图像的过程。尤其是它的非线性调制特性，在生物医学图像处理中具有广泛的应用前景。本文结合PCNN的最新理论研究成果，开展了如下的研究工作：(1)在植物体细胞量化分析研究中，需要计算在不同发育阶段细胞内蛋白质、核酸和淀粉等生物大分子含量和分布的动态变化。整个量化分析过程中，噪声干扰的抑制起到关键性的作用。由于植物细胞切片图像具有细胞和背景相似、灰度差别小的特点，传统方法在滤除图像噪声的同时会导致细胞图像边界模糊。在研究了PCNN时空特性的基础上，本文提出了植物细胞切片图像混合噪声滤除算法。(2)传统图像编码技术的出发点是消除图像数据的统计冗余信息，如信息熵冗余、空间冗余等。不规则区域编码算法，首先使用PCNN将图像分割成人眼敏感的边缘纹理部分和变化缓慢的平坦区域，对这两类采用不同方法编码。仿真实验证明，不规则区域编码技术在高压缩比下保持细节能力仍然良好，解决了以JPEG为代表的块状编码方法带来的失真问题，它是基于视觉特性的图像压缩技术的一个重要发展方向。(3)医学超声图像具有复杂多样、对比度低、斑点噪声多的特点，使得超声图像分析处理工作面临着困难。本文将PCNN引入超声图像处理领域，进行研究探索，提出了基于PCNN的超声图像增强算法和边缘提取算法，为超声图像提供了新的处理工具。更多还原

【Abstract】 Study of modern image processing indicates that the processing techniques must be of high speed, high quality and intelligence, and can also simulate the biologic vision system. "The Third-generation Neural Network"-- Pulse Coupled Neural Network (PCNN) is a new artificial neural network which comes from the research of small mammals’ visual properties, namely, the synchronous pulse burst phenomenon of cat’s visual cortex. PCNN is more similar to the course of vision system processes images because it is the accurate simulation of mammals’ vision system. Especially, its nonlinear modulation characteristic has wide application in biomedical image processing. In this paper, the following works have been done according to the latest research in PCNN:(1) The research of plant body cell needs to account the content of macromolecules such as cellular protein, nucleic acid and starch as well as distributions of enzyme and calcium ion in different develop stages. Noise reduction is a crucial step in the whole quantitative analysis processing. In cell slice image, cell and background are similar, and their intensities are less different. Traditional methods can reduce image noise, but they would blur details at the same time. In this paper, a novel mixed noise removal algorithm for slice images of plant cell based on PCNN is proposed for the first time.(2) Image compression algorithms with effective storage and high quality of restoration are needed to meet large numbers of data and real-time multimedia techniques. The traditional image compression techniques, including Run-length Coding, Huffman Coding, Arithmetic Coding, and so on, make a deep research on decreasing the linear pertinence of images, and aim at reducing the statistical redundancy of originals, such as entropy redundancy and dimensional redundancy. Irregular Regions Image Coding (IRIC) algorithm is presented in this paper. Firstly, the original image is segmented into two kinds of parts by PCNN, one is the contours and textures sensitive to human vision system, and the other is smooth regions. Then they are coded by different methods. Experimental results indicate that IRIC not only can get a high compression ratio, but also protect image details. It is a recommendable technique in modern image compression.(3) Ultrasonic images are usually characterized by complexity, low contrast and all kinds of speckle, making ultrasonic image processing very difficult. PCNN was introduced into the field of ultrasonic image processing, and image enhancement algorithm and edge detection algorithm are proposed in this paper for the first time. Simulations indicate that PCNN is an excellent processing tool for ultrasonic images.更多还原