光学学报, 2018, 38 (9): 0910002, 网络出版: 2019-05-09
协作稀疏字典学习实现单幅图像超分辨率重建 下载: 884次
Collaborative Sparse Dictionary Learning for Reconstruction of Single Image Super Resolution
图 & 表
图 3. (a)目标函数(5)式随迭代次数增加而收敛的情况;(b)聚类数K取不同值时的平均PSNR;(c)字典大小M取不同值时的平均PSNR;(d)参数λ取不同值时的平均PSNR
Fig. 3. (a) Convergence of objective function Eq. (5) along with the increasing of iteration number; (b) average PSNR with different clustering number K; (c) average PSNR with different dictionary size M; (d) average PSNR with different λ
图 4. 不同方法实现第1幅图像超分辨率重建的视觉结果比较。(a)原始HR图像;(b) Bicubic算法;(c) NE+LLE算法;(d) SCSR算法;(e) Zeyde算法;(f) ANR算法;(g) A+算法;(h) SRCNN算法;(i)本文算法
Fig. 4. Visual comparisons of different SR results on No. 1 image using different methods.(a) Original HR image; (b) Bicubic algorithm; (c) NE + LLE algorithm; (d) SCSR algorithm; (e) Zeyde algorithm; (f) ANR algorithm; (g) A+ algorithm; (h) SRCNN algorithm; (i) proposed algorithm
图 5. 不同方法实现第3幅图像超分辨率重建的视觉结果比较。(a)原始HR图像;(b) Bicubic算法;(c) NE+LLE算法;(d) SCSR算法;(e) Zeyde算法;(f) ANR算法;(g) A+算法;(h) SRCNN算法;(i)本文算法
Fig. 5. Visual comparisons of different SR results on No. 3 image using different methods.(a) Original HR image; (b) Bicubic algorithm; (c) NE + LLE algorithm; (d) SCSR algorithm; (e) Zeyde algorithm;(f) ANR algorithm; (g) A+ algorithm; (h) SRCNN algorithm; (i) proposed algorithm
表 1PSNR (dB)与SSIM性能比较
Table1. Performance of magnification in PSNR (dB) and SSIM
|
表 2不同算法的平均重建耗时
Table2. Average running time of different methods
|
邱康, 易本顺, 向勉, 肖进胜. 协作稀疏字典学习实现单幅图像超分辨率重建[J]. 光学学报, 2018, 38(9): 0910002. Kang Qiu, Benshun Yi, Mian Xiang, Jinsheng Xiao. Collaborative Sparse Dictionary Learning for Reconstruction of Single Image Super Resolution[J]. Acta Optica Sinica, 2018, 38(9): 0910002.