Colour Management System:Theory And Application

With the development of technology and the improvement of living quality,imaging devices are becoming more and more popular.However,different devices have their own optical and colourimetric characteristics and thus lead to colour mismatch when communicating colour between devices.Hence,a colour management system is of great importance to ensure a faithful and preferred colour reproduction in colour communication.A complete colour management system,as understood today,includes three essential elements:device characterization,colour appearance model(CAM)and gamut mapping.Firstly,a characterization model is applied to convert the device-dependent RGB space into a device-independent colour space.To achieve the best performance,different characterization models have been developed for different imaging devices,such as the polynomial model for digital camera,the GOG model for CRT display,and the PLCC model for LCD display.In this thesis,all these models will be comprehensively studied and tested for different devices including the state-of-art OLED TV.For OLED,its colourimetric characteristics were initially investigated including channel dependency,primary consistency,channel independency,temporal stability,colour gamut and so on.Afterwards,a new characterization model was built based on the characteristics mentioned above and was proved to give a good prediction performance.The device characterization model describes the relationship between driving signal and its corresponding tristimulus output.However,even for a same tristimulus value,observers will have different perceptions due to the difference of viewing conditions.Hence,different colour appearance models(CAMs)were proposed to accurately describe human perceptions.The objective of a colour appearance model is to precisely quantify human's colour perception.Traditional colour appearance models include CIELAB,CIECAM02 and other CAMs.Those CAMs are usually based on datasets for surface colour,and in standard dynamic range(SDR).Its accuracy for predicting colours on a wide colour gamut(WCG)or a high dynamic range(HDR)display has not been fully studied.Recently,more and more models have been proposed for WCG and HDR displays,including Ictcp model by Dolby Vision and Jzazbz2model by Zhejiang University.In this thesis,all these CAMs,including the most-widely-used CIELAB,the most uniform CAM02-UCS and the newly proposed Jzazbz will be reviewed and their performance will be investigated in the field of gamut mapping.After the discussion of device characterization and colour appearance model,a real example was given to reproduce a real room lighting environment onto a display using a simplified colour management system.Its performance was further verified via a psychological experiment.The results show that using a colour management system could accurately reproduce the real lighting environment and its performance was far better than just using a digital camera.Colour perceptions of an image can be accurately described by the use of device characterization model and colour appearance model.And if the colour rendering capabilities(or say colour gamut)are the same for both the source and the destination displays,a faithful colour reproduction can be easily achieved.However,this is always insatiable.In this condition,a method is needed to avoid such colour mis-matach introduced by the mis-matached colour gamuts.That is to say,colours from the source gamut should be mapped into the destination gamut with least perception change.This is so-called gamut mapping.Gamut mapping usually includes two components,i.e.gamut boundary descriptor(GBD)and gamut mapping algorithms(GMAs).And GMAs can be further divided into 2 different parts,i.e.gamut compression algorithms(GCAs)and gamut extension algorithms(GEAs).Accuracy and smoothness are two elementary requirements for a successful.GBD.The former one ensures an accurate mapping from the source gamut to the destination gamut and the latter one is to avoid sudden change of colour coordinates.In this thesis,different GBDs will be compared and tested for devices,and a method to accurately describe the image boundary will be proposed.Typically,there are two measures to evaluate the performance of GMA.One is accuracy and the other one is preference.Accuracy can be explained as the capability to reproduce a reproduction as close to the original as possible.And preference is regarded as the degree of how observers like the reproduction.In most cases,accuracy is emphasized in gamut compression while preference is more important in the field of gamut extension,as determined by their different purposes.In this study,some most widely adopted GMAs were analysed and compared.Two different GMAs were developed to perform gamut compression and gamut extension based on a newly proposed colour appearance scales,i.e.vividness.For vividness preservation(VP)method,it was designed to preserve as much vividness as possible while for vividness extension(VE)method,it was designed to enlarge the vividness as much as possible.They were further tested with some other state-of-art algorithms,including CIE recommended SGCK and HPMINDE methods.Present results show that these two proposed methods outperformed all the other GMAs and vividness scale was very effective in the field of gamut mapping.Finally,the main contents are concluded and the innovations of this thesis are summarized.Further study perspectives are also proposed.