四元数谱余量彩色图像质量评价 下载: 985次
Color Image Quality Assessment Based on Quaternion Spectral Residual
图 & 表
图 1. RGB图像转换为纯四元数矩阵示意图
Fig. 1. Schematic of color imageconvert to a pure quaternion matrix
下载图片 查看原文
图 2. 彩色图像的四元数梯度图。(a)彩色图像;(b)与(a)相对应的四元数梯度图; (c)彩色图像; (d)与(c)相对应的四元数梯度图
Fig. 2. Quaternion gradient maps of color images. (a) Color images; (b) corresponding quaternion gradient maps of (a); (c) color images; (d) corresponding quaternion gradient maps of (c)
下载图片 查看原文
图 3. QSR-SIM质量评价算法的流程图
Fig. 3. Flowchart of QSR-SIM algorithm
下载图片 查看原文
图 4. TID2013数据库上不同图像质量评价算法的回归曲线。 (a) SSIM;(b) FISM;(c) MS-GMSDc;(d) QSSIM;(e) SR-SIM;(f) QSR-SIM
Fig. 4. Regression curves of different image quality evaluation algorithms in TID2013. (a) SSIM; (b) FISM; (c) MS-GMSDc; (d) QSSIM; (e) SR-SIM; (f) QSR-SIM
下载图片 查看原文
图 5. 参考图像及其不同类型的失真图像。(a)参考图像I07;(b) I07_11_5;(c) I07_12_4;(d) I07_15_3;(e) I07_17_4;(f) I07_24_5
Fig. 5. Reference image and its distortion images of different types. (a) Reference image I07; (b) I07_11_5; (c) I07_12_4; (d) I07_15_3; (e) I07_17_4; (f) I07_24_5
下载图片 查看原文
表 1参考图像I07的5种失真图像在不同评价算法实验中的结果
Table1. Results of five kinds of distorted pictures of reference image I07 in different evaluation algorithm experiments
Algorithm | I07_11_5 | I07_12_4 | I07_15_3 | I07_17_4 | I07_24_5 |
---|
Subjective score | 0.9773 | 2.8636 | 3.0455 | 6.7273 | 0.4762 | QSR-SIM | 0.9701 | 0.9919 | 0.9968 | 0.9976 | 0.9691 | VSI[4] | 0.8964 | 0.9610 | 0.9614 | 0.9868 | 0.8875 | SR-SIM[3] | 0.8283 | 0.9317 | 0.9636 | 0.9803 | 0.8056 | FSIM[5] | 0.7203 | 0.9150 | 0.9610 | 0.9621 | 0.7137 | SSIM[21] | 0.5922 | 0.7208 | 0.9683 | 0.9516 | 0.6019 | VSNR[23] | 8.7561 | 15.0237 | 12.223 | 11.3239 | 8.6929 | PSNR[24] | 23.0007 | 26.1743 | 27.2014 | 25.0836 | 23.0449 | GMSD[25] | 0.2418 | 0.1215 | 0.1291 | 0.0259 | 0.2622 |
|
查看原文
表 2不同评价算法的排名
Table2. Ranking of different evaluation algorithms
Algorithm | I07_11_5 | I07_12_4 | I07_15_3 | I07_17_4 | I07_24_5 |
---|
Subjective score | 4 | 3 | 2 | 1 | 5 | QSR-SIM | 4 | 3 | 2 | 1 | 5 | VSI | 4 | 3 | 2 | 1 | 5 | SR-SIM | 4 | 3 | 2 | 1 | 5 | FSIM | 4 | 3 | 2 | 1 | 5 | SSIM | 5 | 3 | 1 | 2 | 4 | VSNR | 4 | 1 | 2 | 3 | 5 | PSNR | 5 | 2 | 1 | 3 | 4 | GMSD | 4 | 2 | 3 | 1 | 5 |
|
查看原文
表 3算法的时间复杂度比较
Table3. Comparison of time complexity of algorithms
Database | Time /s |
---|
QSSIM | QSR-SIM |
---|
I17_11_2 | 1.1124 | 0.7474 | TID2013 | 520 | 485 |
|
查看原文
表 4各彩色图像质量评价方法的性能比较
Table4. Comparison of various color image quality assessment methods' performance
Database | Parameter | PSNR | SSIM | FSIM | GMSD | MS-GMSD | SR-SIM | QSR-SIM |
---|
TID2013 | SROCC | 0.6869 | 0.7417 | 0.8015 | 0.8030 | 0.8139 | 0.8073 | 0.8169 | KROCC | 0.4958 | 0.5588 | 0.6289 | 0.6352 | 0.6467 | 0.6404 | 0.6268 | PLCC | 0.6748 | 0.7895 | 0.8589 | 0.8575 | 0.8618 | 0.7970 | 0.8413 | RMSE | 0.9149 | 0.7608 | 0.6349 | 0.4762 | 0.6288 | 0.6193 | 0.6701 | CSIQ | SROCC | 0.8058 | 0.8756 | 0.9242 | 0.9515 | 0.9545 | 0.9319 | 0.9435 | KROCC | 0.6084 | 0.6907 | 0.7567 | 0.8021 | 0.8075 | 0.7725 | 0.7821 | PLCC | 0.8000 | 0.8613 | 0.9120 | 0.9453 | 0.9512 | 0.9250 | 0.8049 | RMSE | 0.1575 | 0.1334 | 0.1077 | 0.0856 | 0.0810 | 0.0997 | 0.1618 |
|
查看原文
表 5各评价算法的在TID2013数据库中不同失真类型下的SROCC值
Table5. SROCC values for different distortion types for various evaluation algorithms in the TID2013 database
Database | Distortiontype | MS-SSIM[26] | SSIM_I[27] | SSIM | IFC[28] | VSNR | IW-SSIM | FSIM | SR-SIM | VSI | QSR-SIM |
---|
TID2013 | AGN | 0.8646 | 0.8627 | 0.8671 | 0.6612 | 0.8271 | 0.8438 | 0.8973 | 0.9253 | 0.9460 | 0.9138 | ANC | 0.7730 | 0.7763 | 0.7726 | 0.5352 | 0.7305 | 0.7515 | 0.8208 | 0.8570 | 0.8705 | 0.8111 | SCN | 0.8544 | 0.8505 | 0.8515 | 0.6601 | 0.8013 | 0.8167 | 0.8750 | 0.9225 | 0.9367 | 0.8900 | MN | 0.8073 | 0.7895 | 0.7767 | 0.6932 | 0.7072 | 0.8020 | 0.7944 | 0.7860 | 0.7697 | 0.7036 | HFN | 0.8604 | 0.8688 | 0.8634 | 0.7406 | 0.8455 | 0.8553 | 0.8984 | 0.9132 | 0.9200 | 0.9042 | IN | 0.7629 | 0.7896 | 0.7503 | 0.6408 | 0.7363 | 0.7281 | 0.8072 | 0.8277 | 0.8741 | 0.7885 | QN | 0.8706 | 0.8411 | 0.8657 | 0.6282 | 0.8357 | 0.8468 | 0.8719 | 0.8502 | 0.8748 | 0.8626 | GB | 0.9673 | 0.9724 | 0.9668 | 0.8907 | 0.9470 | 0.9701 | 0.9551 | 0.9620 | 0.9612 | 0.9118 | DEN | 0.9268 | 0.9296 | 0.9254 | 0.7779 | 0.9081 | 0.9152 | 0.9302 | 0.9403 | 0.9484 | 0.9366 | JPEG | 0.9265 | 0.9227 | 0.9200 | 0.8357 | 0.9008 | 0.9187 | 0.9324 | 0.9386 | 0.9541 | 0.9219 | JP2K | 0.9504 | 0.9575 | 0.9468 | 0.9078 | 0.9273 | 0.9506 | 0.9577 | 0.9674 | 0.9706 | 0.9522 | JGTE | 0.8475 | 0.8581 | 0.8493 | 0.7425 | 0.7908 | 0.8388 | 0.8464 | 0.8543 | 0.9216 | 0.7764 | J2TE | 0.8889 | 0.8856 | 0.8828 | 0.7769 | 0.8407 | 0.8656 | 0.8913 | 0.9166 | 0.9228 | 0.8355 | NEPN | 0.7968 | 0.7885 | 0.7821 | 0.5737 | 0.6653 | 0.8011 | 0.7917 | 0.7975 | 0.8060 | 0.8141 | Block | 0.4801 | 0.4563 | 0.5720 | 0.2414 | 0.1771 | 0.3717 | 0.5489 | 0.4731 | 0.1713 | 0.5545 | MS | 0.7906 | 0.7845 | 0.7752 | 0.5522 | 0.4871 | 0.7833 | 0.7531 | 0.6576 | 0.7700 | 0.7762 | CTC | 0.4634 | 0.3800 | 0.3775 | 0.1798 | 0.3320 | 0.4593 | 0.4686 | 0.4705 | 0.4754 | 0.4671 | CCS | 0.4099 | 0.4208 | 0.4141 | 0.4029 | 0.3677 | 0.4196 | 0.2748 | 0.2053 | 0.8100 | 0.5359 | MGN | 0.7786 | 0.8092 | 0.7803 | 0.6143 | 0.7644 | 0.7728 | 0.8469 | 0.8778 | 0.9117 | 0.8110 | CN | 0.8528 | 0.8711 | 0.8566 | 0.8160 | 0.8683 | 0.8762 | 0.9121 | 0.9263 | 0.9243 | 0.9084 | LCNI | 0.9068 | 0.9173 | 0.9057 | 0.8180 | 0.8821 | 0.9037 | 0.9466 | 0.9608 | 0.9564 | 0.9406 | ICQD | 0.8555 | 0.8351 | 0.8542 | 0.6006 | 0.8667 | 0.8401 | 0.8760 | 0.8803 | 0.8839 | 0.8795 | CHA | 0.8784 | 0.8771 | 0.8775 | 0.8210 | 0.8645 | 0.8682 | 0.8715 | 0.8754 | 0.8906 | 0.8494 | SSR | 0.9483 | 0.9488 | 0.9461 | 0.8885 | 0.9339 | 0.9474 | 0.9565 | 0.9614 | 0.9628 | 0.9625 |
|
查看原文
岳靖, 刘国军, 付浩. 四元数谱余量彩色图像质量评价[J]. 激光与光电子学进展, 2019, 56(3): 031009. Jing Yue, Guojun Liu, Hao Fu. Color Image Quality Assessment Based on Quaternion Spectral Residual[J]. Laser & Optoelectronics Progress, 2019, 56(3): 031009.