基于稀疏表示的Retinex图像增强算法研究

发布时间：2018-04-21 08:42

本文选题：Retinex + 图像增强　；参考：《哈尔滨工程大学》2016年硕士论文

【摘要】：图像增强是图像处理的重要环节,其目的对图像进行加工,按照特定的要求改进图像视觉效果或者增强其中一类信息,突出感兴趣的信息或抑制不感兴趣的信息,得到一个效果更好的增强图像。其中,降质图像增强应用最为广泛,而由于恶劣环境导致的降质图像由于其成因的复杂性,特殊性,目前关于降质图像增强的研究都具有一定的局限性,亟需得到改进和加强。因此,本文针对恶劣环境,包括天气、光照、噪声等原因引起的降质图像的机理,基于变分框架的Retinex算法,进行降质图像增强的研究。本文首先介绍了图像增强的原理,并对其中基于Retinex算法的图像增强方法进行了比较,分析其优缺点。以基于变分框架的Retinex算法为基础进行研究。针对其对反射分量的先验知识运用不足,提出了使用稀疏表示理论改进的方法。稀疏表示理论中K-SVD算法对分段定常性质还原性较好的,但是由于字典更新运算量过大导致运算速度缓慢,因此采用批量正交匹配算法,在字典更新过程中只针对更新列进行计算,大大减少运算量并提升了运算速度。并将这种稀疏表示理论应用于变分框架的Retinex算法中,对降质图像进行增强。基于上诉理论,使用本文算法增强了针对不同类型的恶劣环境造成的降质的图像,为证明算法有效性将仿真结果与其他算法进行了比较,证明了本文算法在亮度提升,对比度增强,结构和细节还原上有良好的效果。
[Abstract]:Image enhancement is an important part of image processing. Its purpose is to process the image, to improve the visual effect of the image or to enhance one kind of information according to the specific requirements, to highlight the information of interest or to suppress the information of no interest. Get a better enhancement image. Among them, degraded image enhancement is the most widely used. However, due to the complexity and particularity of the cause of formation, the research on degraded image enhancement has some limitations, which needs to be improved and strengthened. Therefore, aiming at the mechanism of degraded image caused by bad environment, including weather, illumination, noise and so on, this paper studies the enhancement of degraded image based on Retinex algorithm of variational framework. In this paper, the principle of image enhancement is introduced, and the image enhancement methods based on Retinex algorithm are compared, and their advantages and disadvantages are analyzed. The Retinex algorithm based on variational framework is studied. Aiming at the shortage of prior knowledge of reflection component, an improved method using sparse representation theory is proposed. In sparse representation theory, K-SVD algorithm is more effective in reducing piecewise invariability, but because of the slow operation speed due to too much operation of dictionary updating, batch orthogonal matching algorithm is adopted. In the process of dictionary updating, only the update column is calculated, which greatly reduces the computation cost and improves the operation speed. The sparse representation theory is applied to the Retinex algorithm of the variational frame to enhance the degraded image. Based on the appeal theory, this paper uses the algorithm to enhance the degraded image caused by different types of bad environment. In order to prove the effectiveness of the algorithm, the simulation results are compared with other algorithms, and it is proved that the brightness of the algorithm in this paper is improved. Contrast enhancement, structure and detail restoration has a good effect.
【学位授予单位】：哈尔滨工程大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TP391.41

【参考文献】