一种面向颜色校正的拼接图像质量评价方法 
下载： 1449次
A Stitched Image Quality Assessment Method for Color Correction
宁波大学信息科学与工程学院, 浙江 宁波 315211
图 & 表
图 1. 原始拼接图片序列。(a)场景1;(b)场景2;(c)场景6;(d)场景4;(e)场景8;(f)场景3;(g)场景9;(h)场景5;(i)场景7;(j)场景10
Fig. 1. Original stitched image sequence. (a) Scene 1; (b) scene 2; (c) scene 6; (d) scene 4; (e) scene 8; (f) scene 3; (g) scene 9; (h) scene 5; (i) scene 7; (j) scene 10
下载图片 查看原文
图 2. 建库流程图
Fig. 2. Flow chart of building a database
下载图片 查看原文
图 3. 场景2在色差1下的拼接结果。(a)左图像;(b)右图像;(c)算法1;(d)算法2;(e)算法3;(f)算法4;(g)算法5;(h)算法6;(i)算法7;(j)基准图像
Fig. 3. Stitched results of scene 2 in case of color difference 1. (a) Left image; (b) right image; (c) Alg#1; (d) Alg#2; (e) Alg#3; (f) Alg#4; (g) Alg#5; (h) Alg#6; (i) Alg#7; (j) standard image
下载图片 查看原文
图 4. 打分界面
Fig. 4. Scoring interface
下载图片 查看原文
图 5. DMOS值的分布情况
Fig. 5. Distribution of DMOS values
下载图片 查看原文
图 6. 所提方法的流程图
Fig. 6. Flow chart of the proposed method
下载图片 查看原文
图 7. 本文方法的预测值与DMOS的关系曲线
Fig. 7. Relationship between predicted value of the proposed method and DMOS
下载图片 查看原文
表 1颜色校正算法
Table1. Colorcorrection algorithms
| Serial number | Algorithm | Reference |
|---|
| Alg#1 | Brightness function | Method in Ref. [11] | | Alg#2 | Brightness and contrast functions | Method in Ref. [12] | | Alg#3 | Cumulative histogram mapping | Method in Ref. [13] | | Alg#4 | Different color emotion transfer function | Method in Ref. [14] | | Alg#5 | Single color emotion transfer function | Method in Ref. [14] | | Alg#6 | Global color transfer | Method in Ref. [15] | | Alg#7 | Global color transfer in correlated color space | Method in Ref. [16] |
|
查看原文
表 2色差参数
Table2. Color difference parameters
| Number of image | Brightness | Contrast | Saturation |
|---|
| 01-20 | -99 | 0 | 50 | | 21-40 | -50 | 20 | -50 | | 41-60 | 0 | -60 | 50 | | 61-80 | 50 | 0 | 100 | | 81-100 | -30 | -30 | -40 |
|
查看原文
表 3本文方法的PLCC和SROCC值
Table3. PLCC and SROCC values of the proposed method
| Metric | PLCC | SROCC |
|---|
| SRRR | 0.5547 | 0.6076 | | FSIMc | 0.4361 | 0.4633 | | ΔSIFT | 0.1271 | 0.125 | | VSI | 0.1677 | 0.2595 | | CSQIA | 0.6676 | 0.6508 |
|
查看原文
表 4色差2条件下经不同颜色校正算法处理后10种场景的CSIQA分数
Table4. CSIQA scores of 10 scenes processed by different color correction algorithms under color differences condition 2
| Scene | Alg#1 | Alg#2 | Alg#3 | Alg#4 | Alg#5 | Alg#6 | Alg#7 | Alg#8 |
|---|
| Scene 1 | 0.8084 | 0.8093 | 0.5909 | 0.7839Δ | 0.7934 | 0.8151* | 0.7913 | 0.8066 | | Scene 2 | 0.7779 | 0.7460 | 0.7157 | 0.5409 | 0.5047 | 0.8047 | 0.9165* | 0.7854 | | Scene 3 | 0.7176 | 0.7974 | 0.8178Δ | 0.2293 | 0.7211 | 0.9953 | 0.9833Δ | 0.8333 | | Scene 4 | 0.8262 | 0.8353 | 0.4152 | 0.5431 | 0.3020 | 0.7290 | 0.7290 | 0.8006 | | Scene 5 | 0.8315* | 0.6520 | 0.5833 | 0.3972 | 0.4801 | 0.7837 | 0.7325 | 0.7417 | | Scene 6 | 0.7429 | 0.7778 | 0.8017 | 0.5973 | 0.6688 | 0.7801 | 0.8248* | 0.7183 | | Scene 7 | 0.8569 | 0.8220 | 0.5883 | 0.6379 | 0.8185Δ | 0.8531 | 0.7945 | 0.8304 | | Scene 8 | 0.7087 | 0.7000 | 0.7376 | 0.6963 | 0.7294 | 0.8630* | 0.5571 | 0.7155 | | Scene 9 | 0.7132 | 0.7300 | 0.3512 | 0.5880 | 0.6357 | 0.7704 | 0.7938* | 0.7635 | | Scene 10 | 0.6955 | 0.7824* | 0.5332 | 0.7610 | 0.7542 | 0.7551 | 0.4506 | 0.7706 |
|
查看原文
表 5不同场景下不同颜色校正算法与基准算法之间的CSIQA差值
Table5. CSIQA differences between different color correction algorithms and benchmark algorithmsin different scenes
| Scene | Alg#1 | Alg#2 | Alg#3 | Alg#4 | Alg#5 | Alg#6 | Alg#7 | Alg#8 |
|---|
| Scene 1 | 0.0018 | 0.0027 | -0.2157 | -0.0228 | -0.0132 | 0.0084 | -0.0153 | 0.0000 | | Scene 2 | -0.0075 | -0.0394 | -0.0696 | -0.2445 | -0.2807 | 0.0193 | 0.1311 | 0.0000 | | Scene 3 | -0.1157 | -0.0359 | -0.0155 | -0.6039 | -0.1122 | 0.1620 | 0.1500 | 0.0000 | | Scene 4 | 0.0256 | 0.0347 | -0.3854 | -0.2575 | -0.4986 | -0.0716 | -0.0716 | 0.0000 | | Scene 5 | 0.0898 | -0.0897 | -0.1584 | -0.3445 | -0.2616 | 0.0420 | -0.0092 | 0.0000 | | Scene 6 | 0.0247 | 0.0595 | 0.0835 | -0.1209 | -0.0495 | 0.0619 | 0.1066 | 0.0000 | | Scene 7 | 0.0265 | -0.0084 | -0.2422 | -0.1925 | -0.0119 | 0.0227 | -0.0360 | 0.0000 | | Scene 8 | -0.0068 | -0.0156 | 0.0221 | -0.0192 | 0.0139 | 0.1475 | -0.1585 | 0.0000 | | Scene 9 | -0.0503 | -0.0335 | -0.4123 | -0.1755 | -0.1278 | 0.0069 | 0.0303 | 0.0000 | | Scene 10 | -0.0751 | 0.0118 | -0.2374 | -0.0096 | -0.0164 | -0.0155 | -0.3200 | 0.0000 |
|
查看原文
表 6同种场景下不同色差的CSIQA的值
Table6. CSIQA values with different color differences in same scene
| Color difference | Alg#1 | Alg#2 | Alg#3 | Alg#4 | Alg#5 | Alg#6 | Alg#7 | Alg#8 |
|---|
| Color difference 1 | 0.4551 | 0.6052 | 0.7047 | 0.4082 | 0.4276 | 0.9129 | 0.9481 | 0.5635 | | Color difference 2 | 0.7779 | 0.7460 | 0.7157 | 0.5409 | 0.5047 | 0.8047 | 0.9165 | 0.7854 | | Color difference 3 | 0.2942 | 0.4081 | 0.7084 | 0.2924 | 0.4494 | 0.6628 | 0.6890 | 0.5286 | | Color difference 4 | 0.3380 | 0.4059 | 0.6979 | 0.4148 | 0.4062 | 0.7912 | 0.8587 | 0.8304 | | Color difference 5 | 0.4914 | 0.6541 | 0.7082 | 0.6226 | 0.6808 | 0.9061 | 0.8600 | 0.7213 |
|
查看原文
表 7不同色差下CSIQA的差值
Table7. CSIQA differences under different color differences
| Color difference | Alg#1 | Alg#2 | Alg#3 | Alg#4 | Alg#5 | Alg#6 | Alg#7 | Alg#8 |
|---|
| Color difference 1 | -0.1084 | 0.0417 | 0.1412 | -0.1553 | -0.1359 | 0.3494 | 0.3846 | 0.0000 | | Color difference 2 | -0.0075 | -0.0394 | -0.0696 | -0.2445 | -0.2807 | 0.0193 | 0.1311 | 0.0000 | | Color difference 3 | -0.2344 | -0.1205 | 0.1798 | -0.2362 | -0.0793 | 0.1341 | 0.1604 | 0.0000 | | Color difference 4 | -0.4924 | -0.4245 | -0.1325 | -0.4156 | -0.4243 | -0.0392 | 0.0283 | 0.0000 | | Color difference 5 | -0.2299 | -0.0672 | -0.0131 | -0.0987 | -0.0405 | 0.1847 | 0.1387 | 0.0000 |
|
查看原文
齐美玲, 邵枫. 一种面向颜色校正的拼接图像质量评价方法[J]. 激光与光电子学进展, 2019, 56(3): 031102. Meiling Qi, Feng Shao. A Stitched Image Quality Assessment Method for Color Correction[J]. Laser & Optoelectronics Progress, 2019, 56(3): 031102.
PDF全文
