激光与光电子学进展, 2020, 57 (22): 221015, 网络出版: 2020-10-27
基于Gabor变换和组稀疏表示的敦煌壁画修复算法 下载: 919次
Algorithm for Dunhuang Mural Inpainting Based on Gabor Transform and Group Sparse Representation
图 & 表
图 2. 5个尺度8个方向下获得的Gabor特征图
Fig. 2. Gabor characteristic maps obtained under 5 scales and 8 directions
图 3. 随机像素缺失壁画修复效果对比。(a)原始图像;(b)随机像素缺失的待修复图像;(c) CDD算法;(d)文献[ 11]算法;(e)文献[ 12]算法;(f)本文算法
Fig. 3. Comparison among mural inpainting results with random pixel missing. (a) Original image; (b) mural images with random pixel missing; (c) CDD algorithm; (d) algorithm in Ref.[11]; (e) algorithm in Ref.[12]; (f) proposed algorithm
图 4. 人为添加划痕的壁画修复效果对比。(a)原始图像;(b)人为添加划痕的图像;(c) CDD算法;(d) Criminisi算法;(e)文献[
7]算法;(f)文献[
11]算法;(g)文献[
12]算法;(h)本文算法;(i) CDD算法修复区域的局部放大;(j) Criminisi算法修复区域的局部放大;(k)文献[
7]算法修复区域的局部放大;(l)文献[
11]算法修复区域的局部放大;(m)文献[
Fig. 4. Comparison among mural inpainting results with artificial scratch. (a) Original image; (b) image with artificial scratch ; (c) CDD algorithm; (d) Criminisi algorithm; (e) algorithm in Ref.[7]; (f) algorithm in Ref.[11]; (g) algorithm in Ref.[12]; (h) proposed algorithm; (i) local enlargement of repaired area using CDD algorithm; (j) local enlargement of repaired area using Criminisi algorithm; (k) local enlargement of repaired area using algorithm in Ref.[7]; (l) local enlargement of repaired ar
Fig. 4. Comparison among mural inpainting results with artificial scratch. (a) Original image; (b) image with artificial scratch ; (c) CDD algorithm; (d) Criminisi algorithm; (e) algorithm in Ref.[7]; (f) algorithm in Ref.[11]; (g) algorithm in Ref.[12]; (h) proposed algorithm; (i) local enlargement of repaired area using CDD algorithm; (j) local enlargement of repaired area using Criminisi algorithm; (k) local enlargement of repaired area using algorithm in Ref.[7]; (l) local enlargement of repaired ar
图 5. 真实破损的壁画修复效果对比。(a)原始图像;(b)掩模图像;(c) CDD算法;(d) Criminisi算法;(e)文献[ 7]算法;(f)文献[ 11]算法;(g)文献[ 12]算法;(h)本文算法
Fig. 5. Comparison of mural inpainting results with real damage. (a) Original image; (b) mask image; (c) CDD algorithm; (d) Criminisi algorithm; (e) algorithm in Ref.[7]; (f) algorithm in Ref.[11]; (g) algorithm in Ref.[12]; (h) proposed algorithm
图 6. 真实破损壁画的修复效果对比。(a)原始图像;(b)掩模图像;(c) CDD算法;(d) Criminisi算法;(e)文献[ 7]算法;(f)文献[ 11]算法;(g)文献[ 12]算法;(h)本文算法
Fig. 6. Comparison of mural inpainting results with real damage. (a) Original image; (b) mask image; (c) CDD algorithm; (d) Criminisi algorithm; (e) algorithm in Ref.[7]; (f) algorithm in Ref.[11]; (g) algorithm in Ref.[12]; (h) proposed algorithm
表 1随机像素缺失壁画修复结果的比较
Table1. Comparison among mural inpainting results with random pixel missing
|
表 2人为添加划痕的壁画修复结果比较
Table2. Comparison among mural inpainting results with artificial scratch
|
陈永, 陶美风, 艾亚鹏, 陈锦. 基于Gabor变换和组稀疏表示的敦煌壁画修复算法[J]. 激光与光电子学进展, 2020, 57(22): 221015. Yong Chen, Meifeng Tao, Yapeng Ai, Jin Chen. Algorithm for Dunhuang Mural Inpainting Based on Gabor Transform and Group Sparse Representation[J]. Laser & Optoelectronics Progress, 2020, 57(22): 221015.