计算机工程与应用 ›› 2021, Vol. 57 ›› Issue (17): 203-209.DOI: 10.3778/j.issn.1002-8331.2005-0216

• 图形图像处理 • 上一篇    下一篇

改进TV图像去噪模型的全景图像拼接算法

呼亚萍,孔韦韦,李萌,黄翠玲   

  1. 1.西安邮电大学 计算机学院,西安 710121
    2.陕西省网络数据分析与智能处理重点实验室,西安 710121
  • 出版日期:2021-09-01 发布日期:2021-08-30

Improved Panoramic Image Mosaic Algorithm for TV Image Denoising Model

HU Yaping, KONG Weiwei, LI Meng, HUANG Cuiling   

  1. 1.School of Computer Science & Technology, Xi’an University of Posts and Telecommunications, Xi’an 710121, China
    2.Shaanxi Provincial Key Laboratory of Network Data Analysis and Intelligent Processing, Xi’an 710121, China
  • Online:2021-09-01 Published:2021-08-30

摘要:

全景图像生成过程中易受到噪声的影响。针对该问题,提出一种改进的全变分(Total Variation,TV)图像去噪模型应用于全景图像的拼接。对图像去噪过程中建立的泛函函数进行卷积运算,利用卷积操作降低噪声点的灰度值,求解泛函函数所对应的拉格朗日方程极小值,达到图像去噪的效果;采用SIFT(Scale Invariable Feature Transfomation)特征匹配算法对去噪后的图像进行特征提取和匹配;对待拼接图像进行加权融合处理,优化视觉效果。仿真实验结果表明,与经典算法相比,该研究能够较理想地去除图像中的噪声,降低全景图像拼接过程中的干扰,提高视觉效果。

关键词: 图像去噪, 卷积运算, 全变分(TV)图像去噪模型, 尺度不变特征变换(SIFT)特征匹配, 图像拼接, 图像融合

Abstract:

Panoramic image generation process is susceptible to noise. Aiming at this problem, an improved Total Variation(TV) image denoising model is proposed and applied to the Mosaic of panoramic images. Firstly, convolution operation is carried out on the functional function established in the process of image denoising, and the gray value of noise points is reduced by means of convolution operation. Then, the minimum value of Lagrange equation corresponding to the functional function is solved to achieve the effect of image denoising. Secondly, it uses the SIFT(Scale Invariable Feature Transformation) feature matching algorithm for denoising of image feature extraction and matching. Finally, the splicing image is weighted and fused to optimize the visual effect. The simulation results show that compared with the classical algorithm, this research can remove the noise in the image, reduce the interference in the process of panoramic image mosaic, and improve the visual effect.

Key words: image denoising, convolution operation, Total Variation(TV) image denoising model, Scale Invariable Feature Transfomation(SIFT) feature matching, image mosaic, image fusion