基于单层小波变换的视觉传感图像超分辨率重建  

Super Resolution Reconstruction of Vision Sensing Image Based on Single Layer Wavelet Transform

作　　者：张剑[1] 刘鑫[2] ZHANG Jian;LIU Xin(Department of Computer Engineering,ShanXi Polytechnic College,Taiyuan Shanxi 030006,China;school of computing,Central South University,Changsha Hu'nan 410083,China)

机构地区：[1]山西职业技术学院计算机工程系,山西太原030006 [2]中南大学计算机学院,湖南长沙410083

出　　处：《传感技术学报》2022年第1期122-126,共5页Chinese Journal of Sensors and Actuators

基　　金：新型网络与检测控制国家地方联合工程实验室基金项目(GSYSJ20170009)。

摘　　要：针对目前图像重建方法去噪效果不佳,导致重建后图像分辨率较低的问题,提出基于单层小波变换的视觉传感图像超分辨率重建方法。建立低分辨率和高分辨率两种识别空间,分别计算含有噪声干扰区域、正常区域以及信道噪声参数三者间的欧式距离。利用二维平滑函数定义单层小波变换,有效去除视觉传感图像中的噪声,根据多尺度特性对图像中处于边缘微值的分辨率进行具体检测。对所有高分辨率点实行编码,再将编码后的图像系数按照分辨率的高低顺序整理为集合,输出图像完成重建。仿真实验证明,所提方法重建后图像清晰度较高,且结构相似性(Structural Similarity Index Measurement, SSIM)与峰值信噪比(Peak Signal to Noise Ratio, PSNR)的值均高于对比方法,最高值分别为0.95 dB与34.57 dB,说明所提方法的重建效果较好。Aiming at the problem that the denoising effect of current image reconstruction methods is poor, resulting in the low resolution of the reconstructed image, a super-resolution reconstruction method of visual sensing image based on single-layer wavelet transform is proposed. Two recognition spaces of low resolution and high resolution are established to calculate the Euclidean distance among the noise interference region, normal region and channel noise parameters respectively. The single-layer wavelet transform is defined by two-dimensional smoothing function to effectively remove the noise in the visual sensing image, and the resolution of the edge micro value in the image is detected according to the multi-scale characteristics. All high-resolution points are encoded, and then the encoded image coefficients are sorted into sets according to the order of resolution, and the output image is reconstructed. Simulation experiments show that the reconstructed image of the proposed method has high definition, and the values of structural similarity index measurement(SSIM)and peak signal-to-noise ratio(PSNR)are higher than those of the comparison method, with the highest values of 0.95 dB and 34.57 dB respectively, indicating that the reconstruction effect of the proposed method is good.

关 键 词：图像超分辨率重建 视觉传感图像 单层小波变换 分辨率检测 

分 类 号：TP317.4[自动化与计算机技术—计算机软件与理论]