图像降噪和增强对人脸识别系统识别性能的影响

Effect of Image Denoising and Enhancement on Performance of Face Recognition System

目的研究图像降噪和图像增强方法对人脸识别系统识别性能的影响,以期为人脸识别系统应用过程中的图像处理方法选取提供理论指导和技术方案。方法收集33起人像鉴定领域实际案例中的人脸图像素材,研究以高斯滤波和小波变换为代表的图像降噪技术以及具有边缘保持和小波变换特性的单帧图像超分辨率增强技术对人脸识别系统识别性能的影响,并对不同图像处理方法对人脸识别性能的影响进行量化比较分析。结果本文研究的图像降噪技术均显著提高了人脸识别系统的识别准确性,而图像增强技术虽然提高了人脸图像显示效果,但对人脸识别系统的识别性能无正向促进作用。此外,高斯模糊图像处理的图像降噪方法虽然简单,但与本文研究的其他方法比较,其在人脸识别系统识别性能改善方面效果最显著。结论人脸图像质量对人脸识别系统的识别性能具有显著影响,可以通过图像处理技术改善人脸图像质量进而提高人脸识别系统的识别准确性。其中,图像降噪处理可以显著提高人脸识别系统的识别性能,且比图像增强技术更适合于实际人像鉴定应用中的人脸识别系统识别性能增强。

Objective Face recognition system(FRS)is favorite with public security because it can bring forth direct and visible facial images about the involving people.Therefore,the quality of face image is crucial for FRS to play its role effectively.Adoption of image denoising and enhancement can improve image quality,ever becoming the trend those eligible operators pursue.What effect does such an adoption have on the performance of FRS?This study is to explore into the question,with purpose to provide theoretical and technical reference for the selection of image processing approaches to apply FRS into forensic human image identifi cation.Methods Face image materials were collected from 33 real cases about forensic identification of human images.Image denoising was carried out with Gaussian filtering and wavelet transform handling.Single-frame image super-resolution boosting,incorporated with the characteristics of edge-preserving and wavelet transform,was chosen for image enhancement.All the image processing choices adopted above were statistically analyzed to quantitatively assess their effects on the face recognition.Results Two image denoising methods(Gaussian fi ltering and wavelet transform)did improve the recognizing precision of FRS.However,the image enhancement technique(single-frame image super-resolution boosting)had no positive effect on the discriminability of FRS although it meliorated the displaying effect of the selected facial images.Among the image denoising means,Gaussian fi ltering,albeit simple of handling,brought the maximal improved performance into FRS.Conclusions Face image quality is of signifi cant infl uence on the performance of FRS.The image processing could enhance the quality of face image,thereby capable of improving the recognizing precision of FRS.The image denoising methods utilized here could significantly enhance the performance of FRS,with Gaussian fi ltering being more suitable for improving the performance.