图像复原技术论文.docVIP

摘要 运动模糊图像的复原是图像复原中较常见也是较难的一类，在智能交通系统中有着广泛的应用。本文面向车牌识别应用，对运动模糊图像的复原技术进行了系统的研究与实现。 匀速直线运动模糊图像复原的关键在于运动模糊方向和长度的自动鉴别两个方面。将原图像视为各向同性的一阶马尔科夫过程，通过用双线性插值来进行方向微分，实现了运动模糊方向的自动鉴别算法；根据分析模糊图像的频谱图出现黑色条带的原因、条件以及它的精确位置，实现了运动模糊长度自动鉴别算法。 针对复杂成像情况下的运动模糊图像复原工作，着重解决了含噪运动模糊图像和局部运动模糊图像的复原问题；综合应用椒盐噪声检测器和基于带可变正则化参数的径向基神经网络(I也FN)方法，实现了组合滤波器去噪算法，采用改进的局部运动模糊对象提取算法实现局部运动模糊图像的复原。 开发了车牌模糊图像复原系统。该系统对模糊长度和模糊角度均具有较高的鉴别精度，对于含有噪声的运动模糊图像和局部模糊图像进行相应的去噪处理和对局部模糊对象进行提取，并提供参数调整机制以获得最佳的复原效果。自动实现各种类型的运动模糊车牌图像的清晰恢复，复原的效果图可直接应用于后续的车牌识别等工作。 关键词：图像复原，运动模糊，模糊方向，模糊长度，噪声，局部模糊，车牌识别 ABSTRACT The restoration of motion-blurred images is a familiar and also difficult type in image restoration，thus the study of the motionblurred image restoration is of very extensive operation significance．Towards the license plate recognition application，we systemically study and implement the technology of motion-blurred image restoration． The key problem of restoring constantspeed straight-line motion-blurred images lies in the estimation of motion-blurred direction and motion-blurred length．The original image obeys isotropy Markov process with rank one，Can efficiently identify it with high precision via on directional derivation using bilinear interpolation；realizes automatic estimation of motionblurred length；according to the reason and condition of black strips in the spectrum images of motionblurred images and specified the exact positions of black strips，a method to accurately estimate the blurring length of uniform linear motion blurred images is implemented． For the restoration of motion-blurred images in complex imaging environment，this paper emphasizes on the restoration of noisy motionblurred images and partial motion-blurred images，realizes a combined filter using both the salt-andpepper noise detector and radial basis function network approaches，and devises the picking estimation for partial motion-blurred images． we develop a system of motion．blurred license plate images restoration．The system results in preci