| 基于Contourlet变换阈值选取的算法研究 |
投稿时间：2019-08-25 修订日期：2019-08-27  点此下载全文 |
| 引用本文： |
| 摘要点击次数: 94 |
| 全文下载次数: 0 |
|
|
| 中文摘要:目的 针对CT图像在生成或传输时容易受到噪声影响这一难题,提出了基于Contourlet变换结合PCA阈值的去噪方法。 方 法 首先,对噪声图像进行Contourlet变换。然后,对DFB方向滤波后得到的Contourlet系数进行PCA阈值及硬阈值函数去噪。最后,对去噪后的Contourlet系数进行Contourlet逆变换,得到去噪图像。结 果 通过对图像添加噪声并与Contourlet结合六种不同阈值去噪方法进行对比,采用PSNR评估去噪优劣。实验结果表明：贝叶斯阈值方法PSNR值最高PSNR=94.88 dB,然后是PCA阈值PSNR=83.02 dB,最后依次是配套改进阈值PSNR=73.97 dB、3σ阈值PSNR=73.27 dB、统一阈值PSNR=72.33 dB、通用阈值PSNR=71.50 dB。PCA阈值去噪结果整体平滑性较好,视觉效果较好；贝叶斯阈值去噪结果整体平滑性较差,视觉效果不佳。 结 论 通过将Contourlet 变换框架取代传统小波变换框架并结合PCA 阈值应用在图像去噪中,不仅信噪比有所提高,而且图像视觉效果也明显改善。 |
| 中文关键词:CT图像 去噪 Contourlet Wavelet PCA 阈值 |
| |
| Research on Threshold selection based on Contourlet transform |
|
|
| Abstract:Objective In view of the influence of noise on the generation of CT images. A de-noising method based on Contourlet and PCA for images is proposed. Method First, Contourlet transform is applied to noisy images. Then, the Contourlet coefficients obtained after filtering in DFB direction are threshed to denoise by PCA. Finally, PCA inverse, Contourlet inverse transformation of the Contourlet coefficients after denoising is performed to get the denoised image. Results By adding noise to images and comparing with Wavelet combined with Contourlet wavelet six de-noising methods are compared and PSNR is used to evaluate the advantages and disadvantages of de-noising. The results show that Bayes threshold method obtainsSthe highestSSPSNR = 94.88 dB, Then ,the PCA threshold SPSNR = 83.02 dB, Finally, the matching improved threshold PSNR = 73.97 dB, 3σ PSNR = 73.27 dB, Donoho Johnstone threshold PSNR = 72.33 dB, and universal threshold PSNR = 71.50 dB. PCA threshold denoising results have higher smoothness and better visual effect; Bayes threshold denoising results have worse smoothness and poor visual effect. Conclusions By replacing the traditional wavelet transform frame with the Contourlet transform frame and applying it to image denoising with PCA threshold, not only the signal-to-noise ratio is improved, but also the image visual effect is improved obviously. |
| keywords:CT images De-noising Contourlet Wavelet PCA threshold |
| 查看全文 查看/发表评论 下载pdf阅读器 |
|
|
|
位访问者,今日访问
次