Color Calibration Based on KPCA and ANFIS
-
摘要: 提出结合核主成分分析(KPCA)和自适应神经模糊推理系统(ANFIS)的色彩校正(KPCA_ANFIS)算法.首先将数据通过核函数映射到高维空间,再通过KPCA提取主成分,最后通过ANFIS学习达到色彩校正的目的.实验结果验证了ANFIS用于色彩校正的可行性和有效性,KPCA_ANFIS算法的精度和鲁棒性均优于传统ANFIS.对测试数据(训练数据)的平均误差、最大误差和标准差较传统ANFIS分别下降了37%(45%),34%(40%)和35%(40%).Abstract: An algorithm for color calibration was proposed by integrating an ANFIS(adaptive-network-based fuzzy inference system) with KPCA(kernel principal component analysis).The algorithm maps the input data into a higher dimensional feature space with a kernel function first,then extracts principal components of the mapped data by the PCA,and finally implements color calibration by the ANFIS.Experimental results indicate that the proposed algorithm is feasible and effective,and is superior to the conventional ANFIS in both precision and robustness.The average error,maximum error and the error standard deviation regarding the test data(training data) decrease by 37%(45%),34%(40%) and 35%(40%),respectively.
-
Key words:
- color calibration /
- fuzzy logic /
- ANFIS /
- KPCA /
- feature space
-
KANG H R.Printer-related color processing techniques[C]//Proc.of SPIE.Washington:SPIE,1995,2 413:410-419.[2] 乔新亮.面向彩色打印的3D-LUT优化技术[D].硕士学位论文.西安:西安电子科技大学,2002.QIAO Xinliang.The optimization technology of 3D-LUT for color printing[D].Master Dissertation.Xi'an:Xidian University,2002.[3] JOU J M,KUANG S R,CHEN R D.A new efficient fuzzy algorithm for color correction[J].IEEE Transactions on Circuits and Systems,1999,46(6):773-775.[4] JANG J S R.ANFIS:adaptive-network-based fuzzy inference system[J].IEEE Transactions on Systems,Man and Cybernetics,1993,23(3):665-685.[5] 杨延西,刘丁.基于ANFIS的温度传感器非线性校正方法[J].仪器仪表学报,2005,26(5):511-514.YANG Yanxi,LIU Ding.Nonlinear calibration for temperature sensors based on ANFIS[J].Chinese Journal of Scientific Instrument,2005,26(5):511-514.[6] PAVIA A R C,XU Jianwu,PRINCIPE J C.Kernel principal components are maximum entropy projections[C]//Proc.of 6th Int.Conference on Independent Component Analysis and Blind Source Separation.Berlin:Springer-Verlag,2006:846-853.[7] PEREZ-CRUZ F,BOUSQUET O.Kernel methods and their potential use in signal processing[J].IEEE Signal Processing Magazine,2004:57-65.[8] MULLER K R,MIKA S,RATSCH G,et al.An introduction to kernel-based learning algorithms[J].IEEE Transactions on Neural Network,2001,12(2):181-201.[9] 朱龙云,曾平.一种提高3-CCD扫描仪校正精度的新方法[J].计算机工程与应用,2003,27:130-132.ZHU Longyun,ZENG Ping.A new method of improving the calibration precision of 3-CCD scanners[J].Computer Engineering and Applications,2003,27:130-132.[10] JOVANOVIC B B,RELJIN I S,RELJIN I D.Modified ANFIS architecture-improving efficiency of ANFIS technique[C]//Proc.of 7th Seminar on Neural Network Applications in Electrical Engineering.Serbia and Montenegro:Institute of Electrical and Electronics Engineers Inc.,2004:215-220.[11] 张恒喜,郭基联,朱家元,等.小样本多元数据分析方法及应用[M].西安:西北工业大学出版社,2002.[12] 郭茂祖,王亚东,苏晓红,等.基于BP网络的色彩匹配方法研究[J].计算机学报,2000,23(8):819-823.GUO Maozu,WANG Yadong,SU Xiaohong,et al.Research of color matching method based on BP network[J].Chinese J.Computers,2000,23(8):819-823.[13] JANG,J S R.Input selection for ANFIS learning[C]//Proc.of the 1996 5th IEEE international conference on Fuzzy systems.New Orleans:IEEE Press,1996:1 493-1 499.[14] KIM C H,LEE J J.Adaptive network-based fuzzy inference system with pruning[C]//SICE Annual Conference.Tokyo:Soc.of Instrument and Control Eng.,2003:140-143.
点击查看大图
计量
- 文章访问数: 1307
- HTML全文浏览量: 49
- PDF下载量: 477
- 被引次数: 0