Color Coordinat Systemsfor Accurate Color Image Editing Software by uNq5Ch4a

VIEWS: 16 PAGES: 30

									     Color Coordinate Systems for Accurate
         Color Image Editing Software

     Sergey Bezryadin, KWE International Inc, San Francisco, USA
                      Pavel Bourov, UniqueIC’s, Saratov, Russia




International, Inc.


27 June 2006                                                       1
               What if we significantly change contrast?




27 June 2006                                           2
                        PhotoShop gives us a MESS

Over-lightening
and disappearance
of details on the
wall


                                     Chromatic changes
                                     of colors and
                                     disappearance of
Over-darkening                       details on the flowers
and disappearance
of the details in the
shadow




                                     Chromatic changes
                                     of colors on the
                                     flower box



27 June 2006                                          3
               Is it what YOU want to happen …   .




27 June 2006                                     4
               Is it what YOU want to happen with ORIGINAL?




27 June 2006                                              5
               THIS is what you WANT!




27 June 2006                        6
                                                              Is it possible?

Is this result really possible?
    Yes
    And there are TWO ways:
              A lot of operations in PhotoShop by a highly advanced user.
              Use of alternative algorithms and ONE operation.




                            Let the computer do the routine
                             Let the artist to work wonders



27 June 2006                                                                 7
                                                                   Plan

In this presentation, we will…
    Formulate some of the current statements about what a color is,
     and state which of those we use and which we reject in our project.
    Give new definitions for Brightness (B), Chroma (C), and Hue (H).
    Present 3 new color coordinate systems DEF2, Bef and BCH.
    Define main operations and recommend systems in which they
     work the best.
    Show how to prepare a photo with a high dynamic range for
     printing.




27 June 2006                                                             8
                                                 Statements that we accept
   Color is a sensation, which cannot be recorded nor can it be
    reproduced.

   Color is caused by stimulus, some characteristics of which can be
    measured, in order to create another stimulus such that a human (not
    necessarily a dog or a monkey) is not able to distinguish it from the original
    stimulus (both cause the same sensation).

   Stimulus can be described by Tristimulus Values RGB or XYZ and
    for each stimulus a 3-dimentional vector can be associated. All
    stimuli that cause the same sensation are associated with the same
    vector.

                            That is what we ACCEPT



27 June 2006                                                                     9
                                                       Statements that we reject
   Brightness, Chroma, Hue are not measurable.
         If we want to change their value, we need to know what they are and how to
          measure them.

   Definitions of Brightness and Chroma used in Television, HSV
    model, and others.
   Primary Stimulus RGB or XYZ are orthonormal basis.
   Pythagorean Theorem can be used for calculation of a vector’s
    length in any widely-used color coordinate system.



                               That is what we REJECT




27 June 2006                                                                           10
                     Our Definitions of Brightness, Chroma, and Hue
                                  B – Brightness
                                   a norm of the color vector S.
                                  C – Chroma
                      S            an angle between the color vector S
                                   and an axis D.
           B=||S||                     Axis D is a color vector representing Day
                                        Light (for example D65, D55, EE etc.).

                                  H – Hue
                                   an angle between axis E and the
                                   orthogonal projection of the color
                                   vector S on the plane orthogonal to
                                   the axis D.
                                       Axis E - the orthogonal projection of a color
                                        vector, corresponding to some fixed
                                        stimulus (for example, a monochromatic light
                                        with wavelength 700 nm), on the same plane.


27 June 2006                                                                        11
                                                Linear Color Coordinate System
   Linear CCS is the color coordinate system in which each coordinate
    of two stimuli mix is equal to a sum of corresponding coordinates of
    those stimuli.
         CCS is Color Coordinate System.
         Only Linear CCS may be used for image resize because the use of non-linear
          CCS (such as sRGB IEC/4WD 61966-2-1 or CIE L*a*b*) for image resize leads
          to the violation of energy conservation law and results in visual image artifacts.
         CIE XYZ is the primary linear CCS in Colorimetry. There are two standards: CIE
          XYZ 1931 and CIE XYZ 1964, which basis vectors span different subspaces.




27 June 2006                                                                                   12
                                                                   Linear CCS DEF2
   Linear CCS DEF2 is designed based on the CIE 1931 data.
         Digit “2” indicates 2º Standard Colorimetric Observer.

   DEF2 is orthonormal, its design is based on the following restrictions:
         D > 0 and Е = F = 0 for standard Day light D65.
         E > 0 and F = 0 for monochromatic stimulus with 700 nm wavelength.
         F > 0 for yellow stimulus.
         CCS DEF2 is an orthonormal coordinate system according to J. Cohen metrics.
          We have done a research for design different orthonormal coordinate system.
          We tell you about it in next two presentation today.

   The above restrictions uniquely determine coordinate transformation
    between CIE XYZ 1931 and DEF2.




27 June 2006                                                                            13
                  Coordinate Transformation: CIE XYZ 1931 ↔ DEF2
   Coordinate transformation between CIE XYZ 1931 and DEF2 is
    performed though the matrices of transformation.
         XYZ is essentially a non-orthonormal system according to J. Cohen metrics.


                       D   0.2053    0.7125   0.4670   X 
                                                        
                       E    1.8537  1.2797  0.4429    Y 
                       F    0.3655 1.0120  0.6104   Z 
                                                        


                        X   0.6712 0.4955    0.1540   D 
                                                       
                        Y    0.7061 0.0248  0.5223    E 
                        Z   0.7689  0.2556  0.8645   F 
                                                       




27 June 2006                                                                           14
                                                             Plane D = 1
   Plane, where D = 1, is convenient for depicting Gamut of various image
    reproduction devices, for example, for Gamut of sRGB monitor.



                                          sRGB Monitor Gamut

               White Light




27 June 2006                                                                 15
                                                      Plane Y = 1

               White Light

                                       Plane, where Y = 1, is a
               sRGB Monitor Gamut
                                        plane of constant brightness
                                        according to CIE
                                       It is much less convenient
                                        for Gamut representation.
                                       This an additional illustration
                                        of the fact that XYZ is not
                                        orthonormal.




27 June 2006                                                       16
                                                     Chromatic Coordinates (x,y)

   Chromatic coordinates (x,y) and coordinate system Yxy are widely used for
    Gamut depicting and for illustration of some color image transformations.
                               X                                 Y
                       x                                y
                            X Y  Z                          X Y  Z

   We believe that it is very important to preserve chromatic coordinates
    unchanged when manipulating Brightness and/or Contrast.
   All image editing software (as far as we know) does not meet the above
    requirement.

   Introducing similar chromatic coordinates in DEF is not appropriate.
         Coordinates E and F might take negative values.
         There are stimuli for which (D + E + F) = 0.




27 June 2006                                                                  17
                                 CCS Bef and Chromatic Coordinates e & f
   There is an alternative way of defining chromatic coordinates.

                             
                              B  D2  E2  F 2 ,
                             
                             e  E B ,
                             
                              f  F B.
                             
         B is Brightness.
         e and f are chromatic coordinates.
         Defining Vector direction through its interception with a unit sphere has more
          geometrical sense than coordinates of its intersection with any plane.




27 June 2006                                                                               18
                             Chromatic Coordinates e & f


                             sRGB Monitor Gamut

               White Light




27 June 2006                                          19
                                                         Spherical CCS BCH
   Variables B, C, and H, defined earlier, are spherical coordinates related
    with D, E, F through the following equations:


                       B  D2  E 2  F 2

                       D  B  cosC
                      
                       E  B  sin C  cos H
                       F  B  sin C  sin H
                      


   With this definition, Brightness, Chroma and Hue have a clear physical
    meaning.
   This helps to effectively modify image editing algorithms.




27 June 2006                                                                    20
                                                     Main Operations

We will cover 6 main operations:
   Brightness editing
   Contrast editing
   Saturation editing
   Hue editing
   Color to monochrome transformation
   Global dynamic range modification without affecting local dynamic
    range




27 June 2006                                                            21
                                                                 Brightness Editing
   Brightness modification should not affect chromatic coordinates.
   In CCS BCH it might be made as follows:

                        B  f (B)
         f(B) is a non-negative monotone increasing function.


   For example        B  k  B
         k is a positive number.
         In some graphic editors this transformation is named as program exposure
          compensation.




27 June 2006                                                                         22
                                                                   Contrast Editing
   Contrast modification should not affect chromatic coordinates.
   In CCS BCH it might be made as follows:

                        B  f (B, B0 )
         f(B,B0) is a non-negative monotone increasing function of B.

                                             
                                    B 
   For example        B'  Bavg       
                                   B 
                                    avg 
         γ is a positive number.
         Bavg is an average brightness in some neighborhood of point.




27 June 2006                                                                     23
                                                                 Saturation Editing
   Saturation modification should affect neither Brightness, nor Hue.
   In CCS BCH it might be made as follows:

                        C   g (C )
         f(B) is a non-negative monotone increasing function.


   For example         C'=k ·C
         k is a positive number.
         This operation decreases saturation in two times (colors become more faded).




27 June 2006                                                                             24
                                                                  Hue Editing
   Hue modification should affect neither Brightness, nor Saturation.
   In CCS BCH it might be made as follows:

                        H   h(H )
         h(H) is some function.


   Usually, in order to make this transformation, graphic editors apply a turn on
    a fixed angle α.
   In CCS BCH it might be made as follows:

                        H  H 




27 June 2006                                                                    25
                                      Color to Monochrome Transformation
   Color-to-monochrome transformation should not affect Brightness.
   It can be made, for example, as follows:

                        H   H0
                        C   C0
         C0 is Chroma of the chosen color.
         H0 is Hue of the chosen color.
         For grey image, C = H = 0




27 June 2006                                                           26
                                        Global Dynamic Range Modification
   Global dynamic range modification should not affect chromatic
    coordinates.
   Global dynamic range modification should not affect Local dynamic range.
                                              
                                    B0 
   For example           B'  B       
                                   B 
                                    avg 
         γ is a positive number.
         Bavg is an average brightness in some neighborhood of point.
         B0 is a chosen fixed level of brightness.

   This transformation allows for preparation of a photo with a high dynamic
    range for printing (paper has a dynamic range about 30).
         High dynamic range my happen due to big parts of an image being lightened by
          sources of different brightness.
         The brightness of different parts may differ in hundreds or thousands times.


27 June 2006                                                                             27
               Original High-Dynamic Range Photo




27 June 2006                                  28
               Photo Prepared for Printing




27 June 2006                            29
               Thank You!




27 June 2006                30

								
To top