說明 :1、以下表述都是從各論文收集而來,並非原創;
2、由於引文較多,不一一列出,在此對所有作者表示感謝;如有侵權,請聯
系,覈實後刪除!
3、 由於是持續更新,所以每部分的內容沒有排序,可能會有些亂,各位請見
諒!
***************************************************************************************************************************************************************************************
摘要:
算法描述:
實驗設置說明:
1、說明參數如何設置:
For the sake of simplicity we consider here a linear function of the initial noise level, from a= 0 for b= 0 to a= 0.9 for b= 50.
2、客觀評價參數說明(PSNR和SSIM):
We evaluate our denoising results with two image quality measures: the popular Peak Signal to Noise Ratio (PSNR)and the Structural Similarity Index (SSIM) . While simple and practical, the PSNR relies only on the absolute difference pixel by pixel, and does not provide a good signal fidelity measure . As such, its ability to compare images from a human perception point of view is poor. The SSIM is somehow a more complete image quality measure, which builds upon the idea that human perception is highly adaptive to structural information from images and visual scenes 。
實驗結果說明:
1、說明自己的算法整體要優於比較算法,但是不是所有的情況:
in all cases, significant improvements have been made in this objective measure, and the proposed algorithm generally (but nExperimental results justify the performance of the proposed learning-based up-sampling scheme,
which significantly outperforms the state-of-the-art up-sampling algorithms in
terms of PSNR (more than 1 dB improvement), M-SSIM and subjective quality. ot always) provides superior PSNR results relative to the other two algorithms.
2、說明自己算法在客觀指標以及主觀效果都要好
Experimental results justify the performance of the proposed learning-based up-sampling scheme, which significantlyoutperforms the state-of-the-art up-sampling algorithms in terms of PSNR (more than 1 dB improvement), M-SSIM and subjective quality.
總結:
*************************************************************************************************************************************************************************************
作者簡介: 在讀研究生,專注於圖像處理技術研究
**********************************************************************************************************************************************************************************