基于分数阶粒子群的Otsu图像分割算法研究

发布时间：2018-04-19 09:19

本文选题：图像去噪 + 图像增强　；参考：《宁夏大学》2017年硕士论文

【摘要】：图像分割作为图像分析与处理的关键,对图像的边缘目标提取具有重要影响。Otsu分割方法是常用的图像分割方法之一,应用范围广泛。本文针对传统Otsu算法运行时间长、计算复杂且优化Otsu算法的传统粒子群算法收敛速度慢、容易陷入局部最优的缺点,提出了用分数阶微积分算法优化粒子群算法的方案,将Otsu算法、分数阶微积分算法、粒子群算法三种算法进行结合并改进,并将新算法应用于图像分割。本文主要完成的工作和内容如下:(1)在图像分割之前,结合分数阶幅频特性曲线以及分数阶微积分在图像去噪、增强处理中的优越性,针对不同特征的图像不适合用相同的分数阶次来处理,且分数阶次需要人为设定的缺点,提出了一种自适应分数阶微积分的图像去噪、增强算法。根据图像中像素点的纹理、噪声强弱应采用不同分数阶次处理的特点,结合分数阶次在一定范围内能够取得较好效果的特点,提出了不同梯度值下的分数阶次自适应公式,以确保对于梯度值大的噪声强点,取较小的负阶次;对于梯度值小的噪声弱点,取较大的负阶次;对于梯度值大的图像边缘,取较大的正阶次;对于梯度值小的图像纹理,取较小的正阶次。针对不同类型图像,采用自适应阶次的分数阶微分、积分对待分割图像实现去噪、增强预处理。(2)在深入分析传统Otsu算法、分数阶微积分算法及粒子群优化算法理论的基础上,将三种算法结合并改进,提出了分数阶粒子群Otsu图像阈值分割(ImFpsoOtsu)算法。首先采用灰度级-梯度二维直方图算法,以Otsu算法的最大类间方差为适应度函数。然后通过引入粒子进化因子,利用粒子的进化信息自适应更改分数阶次α,同时通过速度增量为零来更新粒子速度、位置值。最后结合传统粒子群粒子更新公式并采用粒子对称分布的改进粒子群算法获取最佳阈值,将目标从图像中分割出来。分数阶粒子群Otsu算法,最终实现了图像的有效分割,解决了传统的粒子群优化算法陷入局部最优的问题,提高了收敛速度。实验结果表明,本文提出的自适应分数阶次的图像预处理算法,从主观视觉和客观的信噪比、熵值上优于传统算法,即图像去噪、增强效果更好。基于分数阶粒子群的Otsu算法,从视觉效果和适应度曲线收敛程度验证了本文算法在保证分割精度的同时,收敛速度更快。
[Abstract]:As the key of image analysis and processing, image segmentation has an important impact on edge target extraction. Otsu segmentation method is one of the commonly used image segmentation methods, and has a wide range of applications.Aiming at the disadvantages of the traditional Otsu algorithm, such as long running time, complex computation and low convergence speed and easy to fall into local optimum, a scheme of particle swarm optimization based on fractional calculus algorithm is proposed in this paper, in which the convergence speed of the traditional particle swarm optimization algorithm is slow and the algorithm is easy to fall into the local optimum.The Otsu algorithm, fractional calculus algorithm and particle swarm optimization algorithm are combined and improved, and the new algorithm is applied to image segmentation.The main work and contents of this paper are as follows: before image segmentation, combining fractional order amplitude-frequency characteristic curve and fractional calculus in image denoising, enhancing the superiority of image processing,An adaptive image denoising and enhancement algorithm based on fractional calculus is proposed to solve the problem that the image with different features is not suitable for processing with the same fractional order and the fractional order needs to be set artificially.According to the texture of pixels in the image, the noise intensity should be processed by different fractional order, combined with the characteristic that fractional order can get better effect in a certain range, a fractional order adaptive formula with different gradient values is put forward.In order to ensure that the negative order is smaller for the noise intensity point with large gradient value, the negative order is larger for the noise weakness with small gradient value, the larger positive order is taken for the image edge with large gradient value, and the image texture with small gradient value is obtained.Take a smaller positive order.According to different types of images, using fractional differential of adaptive order, integral treatment of segmented image to achieve denoising, enhancement preprocessing.) based on the in-depth analysis of the traditional Otsu algorithm, fractional calculus algorithm and particle swarm optimization theory.By combining and improving the three algorithms, a fractional order particle swarm optimization (Otsu) algorithm for threshold segmentation of Otsu images is proposed.Firstly, the gray-grads two-dimensional histogram algorithm is used, and the maximum inter-class variance of the Otsu algorithm is taken as the fitness function.Then, by introducing the particle evolution factor, the fractional order 伪 is changed adaptively by using the evolution information of the particle, and the particle velocity and position value are updated by increasing the velocity to zero.Finally, combined with the traditional particle updating formula and the improved particle swarm optimization algorithm of particle symmetry distribution, the optimal threshold is obtained, and the target is segmented from the image.The fractional-order particle swarm optimization (Otsu) algorithm realizes the effective segmentation of the image, solves the problem that the traditional particle swarm optimization algorithm falls into the local optimum, and improves the convergence speed.The experimental results show that the proposed adaptive fractional order image preprocessing algorithm is superior to the traditional algorithm in terms of subjective vision and objective SNR, I. e., image denoising, and the enhancement effect is better.The Otsu algorithm based on fractional particle swarm optimization verifies that the proposed algorithm can guarantee the segmentation accuracy and converge faster from the visual effect and the convergence degree of fitness curve.
【学位授予单位】：宁夏大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.41

【参考文献】