通信学报

• 学术论文 • 上一篇    下一篇

基于非线性复扩散耦合冲激滤波器的图像放大算法研究

席志红1,海涛1,2   

  1. 1.哈尔滨工程大学 信息与通信工程学院,黑龙江 哈尔滨 150001;2.南阳师范学院 物理与电子工程学院,河南 南阳 473000
  • 出版日期:2014-02-25 发布日期:2014-02-15
  • 基金资助:
    国家自然科学基金资助项目(60875025);河南省科技厅重点科技攻关基金资助项目(112102210326)

Image enlargement research baised on anisotropic complex diffusion coupling to shock filter

  • Online:2014-02-25 Published:2014-02-15

摘要: 为了提高放大算法的适应性,采用改进的非线性复扩散和自适应冲激滤波器,提出了一种图像放大方法。根据像素局部方差进行自适应改变扩散门限,扩散图像的虚部除以扩散时间以消除扩散时间的影响,特别是初期扩散近似线性扩散的特性,得到改进的复扩散模型耦合冲激滤波器进行无噪图像放大。对于噪声图像放大,根据像素局部方差进行自适应非线性复扩散,耦合局部方差约束的冲激滤波器增强模糊的图像边缘和细节。自适应非线性复扩散通过局部方差和图像二阶导数相结合分辨边缘和噪声,对噪声进行平滑的同时保持边缘,克服了复扩散不能分辨噪声和边缘的缺陷,同时保持复扩散保护斜坡结构,免除阶梯效应的优点。仿真实验验证了所提算法不仅对无噪图像有较好的放大效果,而且对一定范围的噪声图像也有较好的放大效果。

Abstract: In order to improve adaptability, the image enlargement method was proposed using improved anisotropic complex diffusion and adaptive Shock filter. For the enlargement of the noiseless image, an improved anisotropic complex diffusion was coupled to shock filter, which adaptively changes the diffusion threshold according to local variance, dividing diffusion time by the diffusion parameter of the image’s imaginary part to remove the diffusion character dependence on the diffusion time, especially the linear diffusion character at the beginning of the diffusion. The noisy image was enlarged by using the adaptive threshold anisotropic complex diffusion filters according to pixel’s local variance coupling with shock filters constrained by pixel’s local variance to enhance the blurred edges. The adaptive threshold anisotropic complex diffusion detects the edge and noise through the combination of local variance and smoothed second derivative of the image, and the proposed method enhances the edge and smoothed the noise, containing the advantage of complex diffusion, as well as preserving the ramp structure and avoiding the staircase effective. The simulations prove prominent performance of the proposed method is not only for noiseless image but also for the certain gauss noisy image.

No Suggested Reading articles found!