基于紧小波框架的多幅盲运动图像修复算法研究

发布时间：2018-06-11 18:18

本文选题：小波框架 + 运动图像修复　；参考：《湘潭大学》2015年硕士论文

【摘要】：在数字图像领域由于运动造成的降质图像的修复问题是一个挑战性问题。在现实生活中随着交通工具和电子设备的日渐发达,对违反交通规则的车辆由于高速行驶被拍下多幅运动模糊的车牌号图像,或者是刑事案件中的犯罪嫌疑人在匆忙经过路边摄像头时被拍下的多幅运动模糊图像等这些普遍存在、运动复杂而又非常重要的图像进行修复有着非常重要的现实意义。这些情况的原始图像和模糊核都未知,修复难度相对较大,属于多幅盲运动图像修复。由于利用多幅盲运动模糊图像进行图像修复时,可以获得较大的图像信息量,能够处理运动复杂的图像修复问题,并能够获得较好的修复效果。因此,多幅盲运动图像的修复算法具有很重要的研究意义。在图像修复领域由于紧小波框架的冗余性能够导致鲁棒性,使得在低精度下获得的图像分解系数能够重构出高精度的清晰图像,在图像修复方面有着很高的重构精度,因此对紧小波框架在图像修复领域的应用有着很大的研究价值。本文结合紧小波框架的优势和图像分类处理的思想提出基于紧小波框架的多幅盲运动分类图像修复方法,该方法对纹理图像和分段光滑图像分别建立图像修复模型,同时在修复模型中考虑了图像中的噪声,加入了噪声约束项。实验表明,该方法既能提高不同类型的运动图像修复质量,又能有效地去除运动图像中的噪声。另外,由于本文提出的基于紧小波框架的多幅盲运动分类图像修复方法中的紧小波框架是固定的,对于不同图像,当其在进行紧小波框架分解时不能得到合适的稀疏度,进而影响图像的修复质量,因此,在后期的研究中本文又提出多幅盲运动分类图像的自适应紧小波框架修复方法,该方法在交替迭代过程中,利用奇异值分解(Single Value Decomposition,SVD)对用来构成紧小波框架分解矩阵的滤波器组进行求解,获得适应于图像的紧小波框架分解矩阵,并将该框架分解矩阵应用于多幅盲运动分类图像修复中。实验表明该方法的图像修复效果有所提高。
[Abstract]:In the field of digital images, the problem of restoration of degraded images caused by motion is a challenging problem. In real life, with the increasing development of vehicles and electronic devices, vehicles that violate traffic regulations are photographed with a number of motion-blurred license plates due to high speed driving. It is very important to repair the motion blur images that the suspects in criminal cases take when they pass the roadside camera in a hurry. These images which are complex and very important have very important practical significance. The original image and blur kernel of these cases are unknown, so the restoration is relatively difficult and belongs to multiple blind moving images. Because of the large amount of image information can be obtained when many blind motion blur images are used, it can deal with the problem of image restoration with complex motion, and can obtain a better effect. Therefore, it is of great significance to study the restoration algorithm of multiple blind moving images. In the field of image restoration, the redundancy of compact wavelet frame can lead to robustness, so that the image decomposition coefficient obtained under low precision can reconstruct high precision clear image, and has high reconstruction accuracy in image restoration. Therefore, the application of compact wavelet framework in the field of image restoration has great research value. Combined with the advantages of compact wavelet framework and the idea of image classification, this paper proposes a new image restoration method based on compact wavelet framework for blind motion classification, which establishes image restoration models for texture images and segmented smooth images, respectively. At the same time, the noise in the image is taken into account in the restoration model, and the noise constraint term is added. Experiments show that the proposed method can not only improve the quality of different kinds of moving images, but also effectively remove the noise from moving images. In addition, because the compact wavelet framework proposed in this paper is fixed in the blind motion classification image restoration method based on compact wavelet framework, for different images, when it is decomposed by compact wavelet frame, it can not get the appropriate sparse degree. Therefore, in the later stage of the research, the adaptive compact wavelet frame restoration method for multiple blind motion classification images is proposed, which is used in the process of alternating iteration. Using singular value decomposition (SVD) single value Decomposition (SVD), the filter banks used to form a compact wavelet frame factorization matrix are solved, and a compact wavelet frame factorization matrix suitable for images is obtained. The frame decomposition matrix is applied to the restoration of several blind motion classification images. Experiments show that the image restoration effect of this method is improved.
【学位授予单位】：湘潭大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TP391.41;O174.2

【相似文献】