摘要
为了更好地实现3维激光扫描图像的去噪,提出一种融合直方图结构相似度(HSSIM)和残差比阈值的改进稀疏去噪算法。利用初始化过完备字典进行稀疏分解,以相似因子代替重构误差作为保真项,利用残差比阈值作为正交匹配追踪算法的迭代终止条件对图像进行去噪,获得了去噪后图像的峰值信噪比及HSSIM指标。结果表明,与基于db2小波变换、多尺度曲波变换和离散余弦变换的去噪方法相比,该算法能获得更好的峰值信噪比和HSSIM数据。在有效去除图像噪声的同时还能更有效地保留图像的细节特征。
In order to get better results of 3-D laser scanning image denoising,an improved sparse representation denoising algorithm was proposed by combining histogram structural similarity（ HSSIM） and residual ratio threshold. The initial overcomplete dictionary was applied in the sparse decomposition. The reconstruction error was replaced by similarity factor as fidelity factor. Then the residual ratio threshold was used as the iteration termination of the orthogonal matching pursuit algorithm to reconstruct the denoised image. Finally,the performance data of denoised image,such as peak signal-to-noise ratio（ PSNR） and HSSIM,were obtained. The experimental results show that the proposed method could provide better PSNR and HSSIM results compared with the image denoising methods using db2 wavelet transform,multiscale curve wave transform and discrete cosine transform. Meanwhile,the structural features can be reserved effectively by the proposed method.
出处
《激光技术》
CAS
CSCD
北大核心
2015年第5期669-673,共5页
Laser Technology
基金
湖南省教育厅科研资助项目(13C122)
湖南省自然科学基金资助项目(13JJ6072)
关键词
图像处理
去噪
直方图结构相似度
残差比
稀疏表示
image processing
de-noising
histogram structural similarity（HSSIM）
residual ratio
sparse representation
