摘要
考虑到在传统的傅里叶频域中星形采样方法的采样区域针对性不强、重建效果不佳等缺点,提出一种基于傅里叶频域中图像显著性信息的变密度压缩采样方法。在传统星形采样的基础上,通过变化密度,在傅里叶频域中显著性区域相对密集采样,同时在非显著性区域相对稀疏采样,以达到更好地恢复图像显著特征信息的目的。实验结果表明,基于相同的信号重构算法,在采样率相同的条件下,该方法重建图像的结构相似度(SSIM)、峰值信噪比(PSNR),以及相对误差(ReErr)均优于传统的星形采样方法。
Considering the shortcomings of traditional star shape sampling method in Fourier frequency domain,for example,the sampling region is not targeted,the quality of reconstruction effect is not well,etc.,we propose a new variable density compressive sampling method which is based on the image salient information in Fourier frequency domain. Based on traditional star shape sampling method,by changing the density we sample relatively densely in saliency region of Fourier frequency domain while relatively sparse in non-saliency region,so as to achieve the goal of better restoring the salient feature information of image. Experimental results show that based on the same signal reconstruction algorithm and under the condition of same sampling rate,the structural similarity（ SSIM）,peak signal-to-noise ratio（ PSNR） and the relative error（ ReErr） of the reconstructed image using the proposed method are all superior to the traditional star shape sampling method.
出处
《计算机应用与软件》
CSCD
2016年第4期164-168,共5页
Computer Applications and Software
基金
国家自然科学基金项目(61171165)
关键词
显著性信息
变密度
傅里叶频域
压缩采样
Saliency information
Variable density
Fourier frequency domain
Compressed sensing
