TITLE OF INVENTION
Wide Color Gamut Anaglyphs
INVENTORS
Monte Jerome Ramstad, a citizen of the U.S.A. residing in Cannon Falls, MN.
CROSS-REFERENCE TO RELATED APPLICATIONS
Not Applicable
STATEMENT REGARDING FEDERALLY SPONSORED RESEARCH OR DEVELOPMENT
Not Applicable
REFERENCE TO SEQUENCE LISTING, A TABLE, OR A COMPUTER PROGRAM LISTING COMPACT DISC APPENDIX
Not Applicable
BACKGROUND
Stereoscopic images generally consist of two images which are related by a small change in the lateral perspective. When viewed through an enabling apparatus, stereoscopic images may provide the perception of stereoscopic depth. Anaglyphs are stereoscopic images wherein different sets of primary colors are used to render the first and second images of the stereo pair. Usually, the spectra of the first and second images do not overlap significantly. Then the first and second images may be viewed selectively
using two complementary color viewing filters. The first viewing filter Fi may be used to view the first image while the second viewing filter F2 may be used to view the second image. The first filter substantially transmits the primary colors of the first image and blocks the primary colors of the second image. The second filter substantially transmits the primary colors of the second image and blocks the primary colors of the first image.
Anaglyphs are often rendered in three primary colors where the first image is rendered in two primary colors while the second image is rendered in one primary color. In red/cyan anaglyphs, the first image is rendered in green and blue primary colors while the second image is rendered in a red primary color. Other types of anaglyphs include blue/yellow and green/magenta anaglyphs. Herein these anaglyphs are called three-color anaglyphs.
Anaglyphs are often used to display stereoscopic images due to their relatively low cost and wide compatibility with display devices. However, conventional anaglyphs have some well known disadvantages. Firstly, conventional anaglyphs generally exhibit a reduced color gamut when viewed through the colored viewing filters. Secondly, conventional anaglyphs generally exhibit retinal rivalry which may cause user discomfort. The prior art contains many methods to improve the color gamut of anaglyphs. The prior art also contains many methods to reduce the retinal rivalry in anaglyphs. However, these anaglyphs still have reduced color gamuts and exhibit retinal rivalry.
It is commonly known that viewing a subject through colored filters may reduce the observed color gamut of the subject. In general, a color filter which transmits only a single primary color may not allow any color hue to be fully perceived through the filter. For example, an image rendered in a pure red primary color may appear to be nearly a grayscale image when viewed through the red filter.
On the other hand, a filter which transmits two primary colors may allow only the hues associated with the two primary colors to be perceived through the filter. The hue
consisting of both primary colors may appear to be nearly a gray color through the filter. For example, a cyan filter (which transmits green and blue light) may allow only blue and green hues or blue and greenish-yellow hues to be perceived through the filter depending on how close the green primary color is to yellow. An image rendered in pure cyan hues may appear to be nearly a grayscale image when viewed through a cyan filter. These phenomena may be confirmed by viewing a digital color spectrum through pure cyan and pure red filters. Software programs for editing digital images often provide a suitable digital color spectrum in their color selection tools.
Since the second image in an anaglyph, is generally perceived as a grayscale image, the color gamut observed in a stereo view of an anaglyph is generally similar to the color gamut of the first image rendered in two primary colors. The first image in an anaglyph generally contributes more to color perception than the second image.
From these observations, one might expect that only blue and yellowish-green hues may be perceived in red/cyan anaglyphs. However, additional color hues are often visible in conventional red/cyan anaglyphs.
One common method of creating red/cyan anaglyphs is to combine the greεn and blue primary channels of the first image with the red primary channel of the second image. This type of anaglyph is often called a "true-color" anaglyph. Surprisingly, red and cyan hues may be perceived in some true-color anaglyphs when viewed through red and cyan viewing filters. In other words, while the single filters do not allow red or cyan hues to be perceived, the stereo view through the two filters may allow red and cyan hues to be perceived. However, the red and cyan hues are generally accompanied by large amounts of retinal rivalry. Similar phenomena occur in analogous blue/yellow and green/magenta true-color anaglyphs.
True-color anaglyphs generally to have too much retinal rivalry for comfortable viewing. Therefore many methods have been developed in the prior art to produce
anaglyphs with less retinal rivalry than observed in true-color anaglyphs. In order to observe less retinal rivalry, anaglyphs are often constructed from images with modified colors. These color modifications may reduce the retinal rivalry observed in the anaglyph, but may also reduce the spectrum or saturation of hues perceived in the anaglyph. Herein these anaglyphs with modified colors and rendered in three primary colors are called partial-color anaglyphs.
There are various editing operations which may be applied to stereoscopic images prior to constructing an anaglyph which are known to reduce retinal rivalry. These include de-saturation of hues and hue substitution. Many methods involve local editing of an image so that the editing functions vary throughout an image. These are very labor intensive and expensive methods to prepare anaglyphs. A particular method of the prior art may cause an average reduction of retinal rivalry in a stereo view while patches of high retinal rivalry remain scattered throughout the stereo view. However, the prior art does not provide a method to reduce the retinal rivalry to arbitrarily low levels for any distribution of initial color content in a stereoscopic image. The prior art lacks a working theory of how to avoid retinal rivalry when producing partial-color anaglyphs.
The conditions which are required to avoid retinal rivalry in color anaglyphs are not described in the prior art. Generally, the prior art contemplates a compromise between the color gamut and the level of retinal rivalry observed in an anaglyph. It is widely believed that retinal rivalry is necessarily present to some degree in color anaglyphs. In order words, it is widely believed that all color anaglyphs have more retinal rivalry than grayscale images. Most efforts of the prior art have been directed toward improving the color gamut of partial-color anaglyphs while accepting a reduced but substantial amount of retinal rivalry.
Methods exist in the prior art to increase the color gamut of anaglyphs by using leaky viewing filters. It is widely known that the range of perceived hues in partial-color
anaglyphs may be expanded to some degree by allowing one or both of the viewing filters to partially transmit or leak a small amount of additional primary colors through the filters. For example, a red filter which also transmits a small amount of green light may allow a dark green hue and an unsaturated red hue to be perceived through the red filter. Or a cyan filter which also transmits a small amount of red light may allow a dark red hue and an unsaturated cyan hue to be perceived through the filter.
Transmitting part of the primary colors of the opposite image through the viewing filters may cause the user to see ghost images or double images in the stereo view. The double images may reduce the ability of the user to fuse the stereo pair and may reduce the perceived stereoscopic depth in the stereo view. Therefore, when using leaky filters, the benefit of the extra hues created by the leak must be balanced against the disadvantage of perceiving less stereoscopic depth.
Conventional cyan filters for viewing red/cyan anaglyphs are often designed to leak a small amount of a red primary color through the filter. This allows a weak reddish hue to be perceived through the cyan filter. However the leaked red primary color creates a ghost of the second image in the view of the first image. Furthermore since the second image is often offset from the first image due to stereoscopic parallax, the red light from the second image is not always at the proper location to contribute correctly to the color of the first image. Similar disadvantages occur when using leaky filters with blue/yellow and green/magenta anaglyphs.
The prior art contains methods to predict the color gamut observed in anaglyphs viewed through leaky filters using conventional color models such as the ClE (International Commission on Illumination) RGB color models. The CIE color models were developed for red, green and blue primary colors. However, it is clear that color perception is drastically changed by color viewing filters. For example, a red viewing filter may change a red color, which is considered a dark color in conventional color models, into a white color which is a bright, unsaturated color. Therefore, applying
conventional color model calculations to predict the color gamut perceivable through color filters has questionable meaning. Furthermore, the color gamut perceivable in an anaglyph depends on the amount and distribution of retinal rivalry. In fact, the effects of retinal rivalry on the perceived color gamut is often greater than the effect of leaking complementary colors through the filters. This is a further reason that color gamut calculations based on conventional color models have limited meaning when applied to conventional anaglyphs.
Grayscale anaglyphs are anaglyphs which are constructed from grayscale versions of stereoscopic images. The grayscale values of the first image are displayed in two primary colors while the grayscale values of the second image are displayed in the remaining primary color. A grayscale anaglyph may appear grayscale when viewed through the anaglyph viewing filters. Grayscale anaglyphs have the advantage of having nearly no perceivable retinal rivalry, but have the disadvantage of not providing colored stereo views. Herein, an anaglyph is considered to be a color anaglyph unless otherwise stated.
BRIEF SUMMARY OF THE INVENTION
One embodiment of the present invention describes a method of creating anaglyphs from stereoscopic images comprising a stereoscopic image including a first image and a second image and the first image represented by two or more color coordinates {Pi,...,Pm} i and the second image represented by two or more color coordinates {Pi,...,Pm}2 and transforming the coordinates {Pi,...,Pm}2 into brightness coordinates {Qi }2 using a brightness transformation GB including brightness transformation data and a first image of an anaglyph represented by the coordinates {Pι,...,Pm} i and a second image of the anaglyph represented the coordinates {Oj }2 whereby the brightness contrast may be balanced in the first and second images of the anaglyph.
An alternative embodiment of the present invention describes a method of creating anaglyphs from stereoscopic images comprising a stereoscopic image including a first image and a second image and the first image represented by two or more color coordinates {Pi,...,Ps} i and the second image represented by two or more color coordinates {Pi,...,Ps}2 and projecting the coordinates {Pi,...,Ps} i into two or more color coordinates {Pi,.. -,Pm} i ' using a projection transformation Gp and projecting the coordinates {P|,...,PS}2 into two or more color coordinates {Pi,...,Pm}2' using the projection transformation Gp, the transformation Gp including projection data and transforming the coordinates {Pi,...,Pm}2' into a brightness coordinate{Qi }2' coordinate using a brightness transformation GB, the transformation GB including brightness transformation data and a first image of an anaglyph represented by coordinates {Pi,...,Pm} i and a second image of the anaglyph represented by coordinates {Qι }2 whereby the brightness contrast may be balanced in the first and second images of the anaglyph.
An alternative embodiment of the present invention describes a method of creating anaglyphs from stereoscopic images comprising a stereoscopic image including a first image and a second image and the first image projectable into two or more color coordinates {Pi,...,Pm} i and the second image projectable into two or more color coordinates {Pi,...,Pm}2 and transforming the coordinates {Pi,...,Pm}2 into brightness coordinates {Qmb using a brightness transformation GB including brightness transformation data and transforming the coordinates {QI B}2 into coordinates {Qi }2 using a rivalry transformation Gz including rivalry data Z5 and a first image of an anaglyph represented by coordinates {Pi,...,Pm} i and a second image of the anaglyph represented by coordinates {Qι }2 whereby the color gamut of a stereo view of the anaglyph with balanced brightness may be expanded with a controlled amount of retinal rivalry.
An alternative embodiment of the present invention describes a method of creating anaglyphs from stereoscopic images comprising a stereoscopic image including a first image and a second image and the first image projectable into two or more color
coordinates {Pi,...,Pm} i and the second image projectable into two or more color coordinates { P ),..., Pm}2 and transforming the coordinates {Pi,...,Pm}2 into masked coordinates (Pi M,- • -,PmM) 2 using a masking transformation GM including a mask image {Mi,...,Mm} and transforming the coordinates (PiMv 5PmMh into brightness coordinates (Qi }i using a brightness transformation GB including brightness transformation data and a first image of an anaglyph represented by coordinates (Pj,...,Pm} 1 and a second image of the anaglyph represented by coordinates (Qi)2 whereby the brightness contrast may be balanced in the first and second images of the anaglyph.
An alternative embodiment of the present invention describes a method of displaying stereoscopic images comprising a stereoscopic image including a first image and a second image and a display apparatus providing three or more primary colors (Pi,...,Pm)and a primary color Qi and the primary colors (Pi,...,Pm) including a red, green, and blue primary color and the spectra of the primary colors (Pi,...,Pm) not substantially overlapping the spectra of the primary color Qi and displaying the first image using the primary colors (Pi,..., P111) and displaying the second image using the primary color Qi.
An alternative embodiment of the present invention describes an apparatus for viewing stereoscopic images comprising a first filter and a second filter and the spectra of the first filter substantially overlapping the spectra of three primary colors (Pi5P25P3) and the spectra of the second filter substantially overlapping the spectra of a primary color Qi and the spectra of the second filter not substantially overlapping the spectra of the first filter.
An alternative embodiment of the present invention describes a method of calibrating an apparatus for displaying stereoscopic images comprising a display apparatus providing two or more primary colors (Pi, ...,P111) and a primary color Qi and a first viewing filter F] substantially transmitting the primary colors (Pi, ...,P111) and blocking the primary color Qi and a second viewing filter F2 substantially transmitting the
primary color Qi and blocking the primary colors {Pi,...,Pm} and an image projectable into two or more color coordinates {Pi,...,Pm} i and transforming the coordinates {Pi,...,Pm} i into brightness coordinates (Qi } i using a brightness transformation GB including brightness transformation data and displaying the coordinates (Pi,...,Pm} ι using the primary colors (Pi,...,Pm} and displaying the coordinates (Qi } i using the primary color Qi and viewing the first image in the primary colors {Pi,...,Pm} through the first filter Fi and viewing the first image in the primary color Qi through the second filter F2 and comparing the brightness contrast of the first image viewed through the first and second filters and selecting new brightness transformation data whereby the brightness transformation data which balances the brightness contrast of anaglyphs may be determined for the display apparatus.
An alternative embodiment of the present invention describes a method of measuring the retinal rivalry in anaglyphs comprising an anaglyph including a first image and a second image and the first image represented by two or more color coordinates (Pi,...,Pm} and the second image represented by a color coordinates Qi and transforming the coordinates {Pi,...,Pm} into brightness coordinates QI B using a brightness transformation GB including brightness transformation data and the retinal rivalry Z a function of Zs, Z = G(Z5) and Z5 = (QI B - QIVQI B for like subject matter in the first and second images.
The present invention is directed to the use of primary colors and special filters to render, display and view digital stereoscopic images in an anaglyph format with a wide color gamut and without retinal rivalry or with a controlled amount of retinal rivalry in the stereo view of the anaglyph.
Some embodiments of the present invention include methods to create and display anaglyphs with little or no perceivable retinal rivalry. The present invention identifies the color property which the visual system uses to process stereoscopic information. Herein this stereoscopic color property is called the brightness contrast or the brightness. The
brightness may be considered a carrier of the brightness contrast in the image. When the brightness contrast is balanced for like subject matter in a stereo view, retinal rivalry may be essentially absent and full stereoscopic depth (true-depth) may be perceived in a stereo view. Herein anaglyphs with balanced brightness contrast are sometimes said to have balanced brightness even though the absolute brightness may be different in the left and right views. Herein an anaglyph with balanced brightness is sometimes called a true- depth anaglyph or a balanced-brightness anaglyph.
Some embodiments of the present invention include methods to create and display anaglyph images using four or more primary colors. Herein, this type of anaglyph is sometimes called a four-color anaglyph even though it may consist of more than four primary colors. A stereo view of a four-color anaglyph may be viewed through viewing filters where a first filter transmits the first image and blocks the second image and a second filter transmits the second image and blocks the first image.
Some embodiments of the present invention include methods to create and display three-color anaglyphs with balanced brightness as well as four-color anaglyphs with balanced brightness. As the retinal rivalry in an anaglyph is reduced, the color gamut observed in the stereo view of an anaglyph may approach the color gamut of the first image viewed through the first viewing filter. Therefore, the color gamut of a balanced- brightness, three-color anaglyph may be approximately a one-dimensional line segment in a CIE (International Commission on Illumination) xy diagram connecting the two primary colors of the first image. The color gamut of a balanced-brightness, three-color anaglyph may contain roughly half the hues in a color wheel. The color gamut of a balanced- brightness, four-color anaglyph may be approximately bounded by a two-dimensional triangle in a CIE xy diagram defined by the three primary colors of the first image. The color gamut of a balanced-brightness, four-color anaglyph may contain roughly all the hues a color wheel.
BRIEF DESCRIPTION OF THE SEVERAL VIEWS OF THE DRAWING
FIG. 1 depicts representative spectra of four primary colors. Also depicted are the spectra of two viewing filters Fi and F2 for different types of four-color anaglyphs of the present invention.
FIG. 2 depicts a representative color gamut of a true-depth anaglyph in which the second image of the anaglyph is rendered in a red Qi primary color and the first image is rendered in red Pi, green P2, and blue P3 primary colors.
FIG. 3 depicts a representative color gamut of a true-depth anaglyph in which the second image of the anaglyph is rendered in a yellow or orange Qi primary color and the first image is rendered in red Pi, green P2, and blue P3 primary colors.
FIG. 4 depicts a representative color gamut of a true-depth anaglyph in which the second image of the anaglyph is rendered in a cyan Qi primary color and the first image is rendered in red Pi, green P2, and blue P3 primary colors.
FIG. 5 depicts a representative color gamut of a true-depth anaglyph in which the second image of the anaglyph is rendered in a blue Qi primary color and the first image is rendered in yellow or red Pi, green P2, and cyan P3 primary colors.
FIG. 6 depicts a representative color gamut of a true-depth anaglyph in which the second image of the anaglyph is rendered in a red Qi primary color and the first image is rendered in green P2, and blue P3 primary colors.
FIG. 7 depicts a representative color gamut of a true-depth anaglyph in which the second image of the anaglyph is rendered in a blue Qi primary color and the first image is rendered in red Pi, and green P2 primary colors.
FlG. 8 depicts a color versus retinal rivalry phase diagram for anaglyphs.
FIG. 9 depicts a method of calibrating the brightness transformation GB in order to create three-color anaglyphs with balanced brightness.
FIG. 10 depicts a method creating three-color, balanced-brightness anaglyphs using a brightness transformation Gβ.
FIG. 11 depicts a method of calibrating the brightness transformation GB in order to create four-color anaglyphs with balanced brightness.
FIG. 12 depicts a method of creating four-color, balanced-brightness anaglyphs using a brightness transformation GB-
FIG. 13 depicts a modified method of creating three-color anaglyphs in order to provide a controlled amount of retinal rivalry using a brightness transformation GB and selected retinal rivalry Zs.
FIG. 14 depicts a representative color gamut of an anaglyph with a selected amount of retinal rivalry in which the second image of the anaglyph is rendered in a red Qi primary color and the first image is rendered in green P2, and blue P3 primary colors.
FIG. 15 depicts a representative color gamut of a true-depth anaglyph in which the second image of the anaglyph is rendered in a red Qi primary color and the first image is rendered in yellow Pi, green P2, and blue P3 primary colors.
DETAILED DESCRIPTION OF THE INVENTION
Four-Color Anaglyphs
Some embodiments of the present invention include methods to display anaglyphs using four or more primary colors. The first image may be rendered in three or more independent primary colors {Pi,...,Pm}, while the second image may be rendered in one primary color Qi. Herein this type of anaglyph will be called a four-color anaglyph even though it may consist of more than four primary colors. Herein the primary colors {Pi, ...,Pm} consist of independent primary colors unless otherwise stated. A dependent primary color is substantially equivalent to a combination of the other primary colors. For example in CMYK printing primary colors, black K is a dependent primary color because black may be obtained from a combination of cyan C, magenta M, and yellow Y primary colors. Subtractive primary colors may be transformed into additive primary color coordinates and are included in the scope of present invention. Dependent primary colors are also included in the scope of the present invention.
In some embodiments of the present invention a pair of viewing filters is provided for viewing the four-color anaglyphs. A first viewing filter Fi substantially transmits the primary colors {Pi,...,Pm} and blocks the primary color Qi. The second viewing filter F2 substantially transmits the primary color Qi and blocks the primary colors {Pi,...,Pm}. It is within the scope of the present invention for the first ahd/or second filters to leak a small amount of the luminance of the blocked primary colors through the viewing filters.
In some embodiments of the present invention, the spectrum of the primary color Qi does not substantially overlap the spectra of the primary colors {P|,...,Pm}. This generally allows the first and second images of the anaglyph to be selectively viewed through first and second color filters respectively.
In some embodiments of the present invention, the spectra of the primary color Qi may overlap some of the spectra of the primary colors {P|,...,Pm} or be equal to the spectra of one of the primary colors Pj. In this embodiment, the polarization state of the primary colors {Pi,...,Pm} may be opposite to the polarization state of the primary color Qi . Then, the first and second images may be selectively viewed using first F] and
second F2 polarized filters where filter Fi transmits light having the polarization state of the primary colors {Pi,...,Pm} and where filter F2 transmits light having the polarization state of the primary color Qi.
In some embodiments of the present invention, the first and second images of an anaglyph may be superimposed on a surface. In other embodiments of the present invention, the first image may be displayed on a first region of a first surface and the second image may be displayed on a second region of a second surface where the first and second regions do not overlap. Examples are: two images displayed side-by-side; or two images displayed in a head mounted display where each eye may view a separate display surface.
The color gamut observed in the stereo view of an anaglyph may approach the color gamut of the first image viewed through the first viewing filter as the retinal rivalry approaches zero. The color gamut of the first image of a four-color anaglyph viewed through the first filter may be a two-dimensional triangle in a CIE xy diagram. Herein the color gamut of an image is considered to be the color gamut of the set of primary colors used to render the image. The color gamut of a set of primary colors is the set of colors which may be rendered using the set of primary colors. The color gamut of a set of primary colors may be depicted by plotting the xy chromacity coordinates of the primary colors on a CIE chromacity diagram, b A property of a CIE chromacity diagram is that the color gamut of a set of primary colors can be depicted by connecting the points representing the primary colors with line segments in the chromacity diagram. The color gamut includes the area bounded by the line segments.
In some embodiments of the present invention, anaglyphs may be displayed by digital display devices including digital projector, LCD displays, and CRT displays. In other embodiments, the anaglyphs may be displayed on print media including opaque and transparent print media.
Figure 1 depicts non-overlapping spectra 102, 104, 106, and 108 representing four primary colors. The first spectra 102 may be a blue primary color. The second spectra 104 may be a cyan or green primary color. The third spectra 106 may be a green, yellow, orange or red primary color. The fourth spectra 108 may be a red or far-red primary color. The primary color Qi may be represented by any of the four spectra in different embodiments of the present invention. In other words, primary color Qi may be blue, cyan, green, yellow, orange, red or far-red. The primary colors (Pi3P?, P3) may be represented by the other three spectra which generally may include red, green, and blue primary colors, but may include yellow, green and blue primary colors, or orange, green, and blue primary colors, or red, green and cyan primary colors. The first viewing filter F] of each embodiment transmits the primary colors {Pi,P2,P3} while the second viewing filter F2 transmits the primary color Qi.
In one embodiment of the present invention, the primary color Qi may be red. Figure 1 depicts the transmission spectra of filters Fi and F2 in plot 118. The red primary color Qi is represented by the fourth spectra 108. The primary color Pi may be red, orange or yellow and is represented by the third spectra 106. Primary color P2 may be a green and is represented by the second spectra 104. Primary color P3 may be a blue and is represented by the first spectra 102. Figure 2 depicts the color gamut 202 of an anaglyphs of the present embodiment in a CIE xy diagram where the primary color Pi is red. The color gamut of the anaglyphs may be represented by the triangle formed by the points Pi, P2, and P3 representing the primary colors of the first image of the anaglyph. The point Qi represents the primary color of the second image. The point Wp represents the conventional white point.
In another embodiment, the primary color Pi may be yellow or orange. Figure 15 depicts the color gamut 1502 of an anaglyph of the present embodiment in a CIE xy diagram where the primary color Pi is yellow. The color gamut of the anaglyphs may be represented bythe triangle formed by the points Pi, P2, and P3 representing the primary colors of the first image of the anaglyph. The point Qi represents the primary color of the
second image. The point Wp represents the conventional white point. The point Wp' represents the approximate location of the apparent white point viewed through the first viewing filter. This embodiment may include images printed in high-fidelity inks which contain four or more independent ink spectra. For example, high-fidelity inks may include a cyan, magenta, yellow, green and orange primary colors.
In another embodiment of the present invention, the primary color Qi may be yellow or orange. Figure 1 depicts the transmission spectra of filters Fi and F2 in plot 116. The yellow or orange primary color Qi is represented by the third spectra 106. The primary color Pi may be red and is represented by the fourth spectra 108. Primary color P2 may be green and is represented by the second spectra 104. The primary color P3 may be blue and is represented by the first spectra 102. Figure 3 depicts the color gamut 302 of an anaglyph of the present embodiment in a CIE xy diagram. The color gamut of the anaglyphs may be represented by the triangle formed by the points Pi, P2. and P3 representing the primary colors of the first image of the anaglyph. The point Qi represents the primary color of the second image. The point Wp represents the conventional white point.
In another embodiment of the present invention, the primary color Qi may be cyan. Figure 1 depicts the transmission spectra of filters Fi and F2 in plot 114. The cyan primary color Qi is represented by the second spectra 104. The primary color P] may be red and is represented by the fourth spectra 108. Primary color P2 may be green and is represented by the third spectra 106. Primary color P3 may be blue and is represented by the first spectra 102. Figure 4 depicts the color gamut 402 of an anaglyph of the present embodiment in a CIE xy diagram. The color gamut of the anaglyphs may be represented by the triangle formed by the points Pi, P2, and P3 representing the primary colors of the first image of the anaglyph. The point Qi represents the primary color of the second image. The point Wp represents the conventional white point.
In another embodiment of the present invention, the primary color Qi may be blue. Figure 1 depicts the transmission spectra of filters Fi and F2 in plot 112. The blue primary color Qi is represented by the first spectra 102. The primary color Pi may be red and is represented by the fourth spectra 108. Primary color P2 may be green and is represented by the second spectra 106. Primary color P3 may be cyan and is represented by the second spectra 104. Figure 5 depicts the color gamut 502 of an anaglyph of the present embodiment in a CIE xy diagram. The color gamut of the anaglyphs may be represented by the triangle formed by the points Pi, P2, and P3 representing the primary colors of the first image of the anaglyph. The point Qi represents the primary color of the second image. The point Wp represents the conventional white point.
When viewing an anaglyph through filters Fi and F2, the white point Wp' generally moves close to the middle of the color gamut of the anaglyph. The shift in the white point may affect the saturation and hues of some colors. In the case of a three-color anaglyph, the white point Wp' may shift toward the center of the gamut line segment. In this case, the hues near the midpoint point of the gamut may be de-saturated or loose color. In the case of a four-color anaglyph, the white point may move toward the center of the triangular gamut. This shift in the white point may allow a full spectrum of hues including red, yellow, green, cyan, blue, and magenta hues to be visible in the color gamut of the anaglyph.
The color gamut of the first image of a three-color anaglyph may be a one- dimensional line segment. Figure 6 depicts the color gamut 602 of a red/cyan anaglyph without retinal rivalryin a CIE xy diagram. The color gamut is the line segment connecting the points Pi and P2 representing the primary colors of the first image of the anaglyph. The point Qi represents the primary color of the second image. The point Wp represents the conventional white point.
Figure 7 depicts the color gamut 702 of a blue/yellow anaglyph without retinal rivalry in a CIE xy diagram. The color gamut is the line segment connecting the points Pi
and P2 representing the primary colors of the first image of the anaglyph. The point Qi represents the primary color of the second image. The point Wp represents the conventional white point.
For a three-color anaglyph, the white point generally moves close to the midpoint of color gamut line segment. Figures 6 and 7 show the approximate location of the white point Wp' viewed through the first color filter. The shift in the white point generally allows about three hues to be visible in the color gamut. For example the color gamut of a red/cyan anaglyph may include yellow, partial green, and blue hues and possibly a cyan tint to the white point. The color gamut of a green/magenta anaglyph may include red and blue hues and possibly a magenta tint to the white point. The color gamut of a blue/yellow anaglyph may include red and green hues and possibly a yellow tint to the white point.
Since the color gamut of a four-color anaglyphs may be two-dimensional, four- color anaglyphs generally may have greater color gamuts than three-color anaglyphs.
Balanced-Brightness Anaglyphs
Retinal rivalry may be associated with unbalanced brightness contrast in the first and second images of a stereo pair. For example, a red object, which may appear relatively bright through a red filter and may appear relatively dark through a cyan filter, may cause retinal rivalry in a stereo view. Generally the human visual system may not easily fuse a dark object observed by one eye with a bright object observed by the other eye. Instead, the user essentially experiences double vision, where the left and right eyes "see" independently, and stereoscopic perception is much reduced or absent. (Stereoscopic perception requires the cooperation of both eyes.)
Similarly, stereoscopic perception may be reduced wherever retinal rivalry is observed in a stereo view. It is a discovery of the present invention that avoiding retinal
rivalry in a stereo view is essentially identical to the condition required for full stereoscopic depth perception. Since the perception of stereoscopic depth depends on observing relationships between objects, retinal rivalry in a small portion of an image may dramatically reduce the perception of stereoscopic depth in other portions of the image as well.
Some embodiments of the present invention include methods for creating anaglyphs with little or no perceivable retinal rivalry. The present invention identifies the color property which is substantially used by the visual system to process stereoscopic information. Herein this stereoscopic color property is called brightness contrast or brightness. The brightness may be considered a carrier of the brightness contrast in the image. When the brightness contrast is balanced in both eyes for like subject matter, retinal rivalry may be essentially absent from the stereo view and full stereoscopic depth (true-depth) may be perceived in a stereo view. The methods of the present invention to create anaglyphs with balanced brightness may be applied to three-color anaglyphs as well as to four-color anaglyphs. Herein anaglyphs with balanced brightness contrast are sometimes said to have balanced brightness even though the absolute brightness may be different in the left and right views. Herein anaglyphs with balanced brightness are sometimes called true-depth anaglyphs or balanced-brightness anaglyphs.
In the prior art, the term brightness is generally used to describe an observed property of light rather than a physical property of light. In the present invention, brightness also describes an observed property of light. However, in the present invention, brightness is defined by the absence of retinal rivalry which may be determined by observation.
Anaglyphs often have about 3-10 times more luminance in the first image than in the second image. This is because the single primary color Qi of the second image does not provide as much luminance as the two or more primary colors of the first image. When the second image has a luminance of about 15-30% of the first image, the second
image may appear to be about the same luminance as the first image due to vision adaptations. The lower luminance of the second image of about 15-30% allows the color gamut in an anaglyph to be similar to the color gamut of the first image.
When the brightness is balanced in an anaglyph, the color gamut of the stereo view may be nearly identical to the color gamut of the first image. For example, when the brightness is balanced, the color gamut of a red/cyan anaglyph may be nearly equal to the color gamut observed in the first image viewed through the cyan filter. Similarly, the color gamut observed in a four-color anaglyph may be nearly equal to the color gamut of the first image. Therefore figures 2-7 and 15 show approximately the color gamuts of balanced-brightness, four-color and three-color anaglyphs. Three-color anaglyphs with balanced brightness generally may have one-dimensional color gamuts depicted schematically in figures 6 and 7. Four-color anaglyphs with balanced brightness generally may have two-dimensional color gamuts depicted schematically in figures 2-5.
The primaiy color Qi may sometimes cause white colors in an stereo view to be shifted toward the hue of the primary color Qi. For this reason, it may be beneficial to perform corrective color processing of the first image to shift the white point of the first image away from the hue of the primary color Qi .
Anaglyphs with substantial retinal rivalry generally may have a greater color gamut than if the retinal rivalry were absent, but then depth perception is reduced. Figure 8 depicts the relationship of the anaglyphs with balanced brightness of the present invention to the three-color anaglyphs without balanced brightness of the prior art in a color versus retinal rivalry phase diagram. The vertical line segment at zero retinal rivalry represents some of the anaglyphs of the present invention. A balanced-brightness, three-color anaglyph of the present invention is drawn at the 50% color point 802 because about half of the hues of a color wheel may be observed in the three-color anaglyphs. For example in a red/cyan anaglyph with balanced brightness, three hues out of six may be perceived: (1) blue, (2) unsaturated green, and (3) near yellow. The missing hues may be
(1) red, (2) magenta, and (3) cyan (although an unsaturated cyan may be perceivable). A balanced-brightness, four-color anaglyph of the present invention is drawn at the 100% color point 804 because a full spectrum of hues may be observed in four-color anaglyphs.
The horizontal line at zero color in figure 8 represents grayscale anaglyphs of the prior art. Grayscale images may produce nearly balanced-brightness anaglyphs without the need for determining the brightness color property of the present invention which the visual system uses to perceive stereoscopic depth. In order to balance the brightness of color anaglyphs, it is necessary to determine how the brightness contrast depends on a color's hue, saturation, and luminance (or luminance contrast). Grayscale images have a constant hue and saturation so the brightness only depends on the luminance (or luminance contrast). Since the luminance (or luminance contrast) may be nearly balanced in a grayscale image, the brightness contrast may be automatically nearly balanced also. Therefore in figure 8, the point 808 at zero color and zero retinal rivalry (0,0) represents most grayscale anaglyphs of the prior art.
In color anaglyphs of the prior art, as the color saturation increases from zero, the retinal rivalry becomes non-zero. In figure 8, curve 818 represents schematically the envelope of prior art anaglyphs which attempt a compromise between retinal rivalry and the color gamut. A true-color anaglyph of the prior art is drawn at the 100% color, 100% retinal rivalry point 806 (100,100) (although a true-color anaglyph might not strictly have 100% color). For example, a pure red or cyan hue in a true-color anaglyph has maximum retinal rivalry. True-color anaglyphs are often described as having full-color.
De-saturating a stereoscopic image will typically reduce the retinal rivalry in the stereo view of the anaglyph compared with a true-color anaglyph. A de-saturation process of the prior art is represented by arrow 810 in figure 8. Methods in the prior art which attempt to maximize the color gamut in the presence of substantial retinal rivalry, or seek a compromise between retinal rivalry and color gamut may be represented by arrows 814 and 816 in figure 8.
Figure 8 depicts how the anaglyphs of the prior art generally lie in a continuum on a color-gamut/retinal-rivalry phase diagram whereas there is a gap between the balanced- brightness anaglyphs of the present invention and the anaglyphs of the prior art. The region 820 between curve 818 and the zero-retinal-rivalry line represents additional methods of the present invention in which the retinal rivalry in an anaglyph may be constrained below certain levels.
Methods
The color property used by the visual system to process stereoscopic depth was discovered by the present inventor through a series of approximation methods while attempting to further reduce retinal rivalry in red/cyan anaglyphs. The methods below are conceptually similar, but progressively better account for the non-linearity of a display device. Method 2 below is the first method described which partially takes into account the non-linearity of the display device. Each method provides control over the retinal rivalry in an anaglyph where the retinal rivalry may be constrained below a certain level by the method. Herein anaglyphs in which the retinal rivalry is constrained below a certain level are called balanced-brightness anaglyphs.
First, it was noticed that retinal rivalry occurs in anaglyphs where an edge in the subject matter is observed with more contrast in one eye than in the other eye. It seemed apparent that balancing the contrast in all subject matter would eliminate retinal rivalry. Secondly, it was noticed that stereoscopic fusion depends largely on the edges of objects being observed by both eyes. Here an edge may be generalized to any type of contrast in an image. In order for all edges to be detected by both eyes for all subject matter, it is apparently sufficient to observe balanced contrast at all brightness levels in the image. Herein the definition of balanced contrast in the first and second views is: the relative contrast in the two views which eliminates retinal rivalry in the stereo view. Then the goal may be to balance the contrast in all subject matter observed by the left and right
eyes in a stereo view in order to eliminate retinal rivalry. Obtaining full-depth perception is an additional benefit of these methods which is apparently linked to the absence of retinal rivalry.
The brightness of a hue is generally an integral of the brightness contrast. Therefore, the present invention provides a definition of brightness which is different from that of the prior art. Herein the brightness is an observed property of light rather than a physical property of light. The relative brightness of different hues depends on subjective evaluation by users. The relative brightness may vary from user to user and may vary over time for the same user. The relative brightness may also depend on the size of a patch of color, the surrounding colors, and on the vision accommodation of the user. Therefore, implementing the methods below may sometimes require skill, patience, training and multiple observers.
Comparing the contrast in the first and second views of anaglyphs may be carried out by using one eye to view in rapid succession the first and second images through the first and second filters respectively. Using one eye reduces some of the problems of eye accommodation and eye variations. Furthermore, the test anaglyphs may be prepared from identical left and right images so that the stereoscopic parallax may be zero throughout the anaglyphs. This also eliminates the issue of variations of subject matter between the first and second images.
Some of the methods described below require the determination of certain experimental parameters. These parameters may be determined for test anaglyphs in which the first and second image are identical. When the first and second images are identical, the color coordinates are identical and there is no need to distinguish between the first and second images' color coordinates. In order to simplify the descriptions of the methods, the descriptions below sometimes may not distinguish between the color coordinates of first and second images. In general, the methods contain a set of color operations, which may include a projection, that may be applied to both images before the
anaglyph is constructed. Then in order to create a balanced-brightness anaglyph, the brightness coordinate may be determined for the second image. Then the anaglyph may be constructed from the modified first image and the brightness coordinate determined from the modified second image. Some of the methods are illustrated below for red/cyan anaglyphs. Adapting these methods to other three-color anaglyphs such as blue/yellow and green/magenta anaglyphs and to four-color anaglyphs are within the scope of the present invention.
Balanced Brightness Method 1
In method 1 of the present invention to create balanced-brightness anaglyphs, the color hues which generally have balanced brightness when viewed through the red and cyan filters may be identified. This may involve comparing the relative brightness of different hues observed through different colored filters. Comparing the brightness of different hues may be complicated by the properties of vision. The relative brightness of different hues such as green and red hues relies on subjective judgements. These judgements may depend on the user, the surrounding colors, the ambient light, the size of a color patch, the hue and luminance accommodation of the eye, and may vary for a user over time. Nevertheless, for a red/cyan anaglyph, it was determined for a particular computer monitor that two hues, a certain greenish-yellow hue and a certain bluish-purple hue, generally balance the brightness observed through a particular pair of red and cyan filters. Therefore, when the colors in a stereoscopic image are projected into these two hues, the observed retinal rivalry may be constrained below a certain level over the entire stereo view.
In terms of the RGB coordinates, the two hues may be summarized as follows:
R = Ot]G greenish-yellow, R = Ot2B bluish-magenta
where cci <1, α2 < 1 may be measurable parameters which determine the hues. To simplify the discussion, but without loss of generality αi + α2 =1 may be assumed.
By subtracting off the gray component W of an image, the αi and α2 may be optimized with more flexibility or accuracy for a particular image. Also a color projection step, discussed below, may be performed with more flexibility with the gray component subtracted off. First, three color hues red r, green g, and blue b may be defined as follows:
r = R-W , g = G-W , b = B-W.
W = min(R,G,B)
Then, the balanced brightness condition may be written as follows:
r' = αig + α2b,
where RA, GA and BA are the primary coordinate values of the anaglyph, r' represents the new coordinate values that replace the original r coordinate values in the initial image. The brightness condition above includes the case of a grayscale image (RA,GA,GA)=(W,W,W) having balanced brightness.
In general, different display media may have different values of αi and α2 which may or may not correspond to different observed hues. For one particular CRT computer monitor, αi was about 0.9 and α2 was about 0.4. For one particular inkjet printer, αi was
about 0.7 and α2 was about 0.1. αi and α2 may be expected to vary greatly for display media which do not conform to certain color standards.
In creating, a red/cyan anaglyph with balanced brightness, the red color coordinate R is replaced with a brightness coordinate. In order to preserve the information in the original R color coordinate, it may be beneficial to first project the original RGB coordinates into new G' and B' coordinates.
This may be implemented by first projecting the (r,g,b,W) color coordinates into new (g,b,W)' coordinates, then calculating the new (r)' color coordinate from the (g, b,W)' coordinates, and finally creating the anaglyph (R,G,B)A:
Red/Cyan anaglyph
Step l : (R,G,B)i -» (r,g,b,W)i5
Step 2: (r,g,b,W), -> (g,b,W)r,
Step 3: (g,b,W)2' -> (r)2' where r = α,g + α2b ,
Step 4: (r.g.b.Wy -» (R5G5B)1' ,
Step 5: (R)2' ,(G5B)1 ' -* (R,G,B)A
where i may be the first or second image.
Step 2 may be implemented by a method which projects the information in the r coordinate into the (g',b',W) coordinates such as:
g' = β.g + β2r, b' = β3b + β4r, and W = β5W + β6r.
where the β, may be selectable parameters. Many other projection methods are possible and are within the scope of the present invention. Herein, the βj parameters are called
projection data. The projection data may be selected by a user according to a user's preferences. Projection data may be selected which substantially transforms the relative brightness of hues in the initial image into the projected image.
Step 5 shows how to construct the anaglyph from RGB coordinates of the first and second images. Steps 1-4 may be applied to the primary colors of the second image to obtain (R)2' . Steps 1 -2 and 4 may be applied to the primary colors of the first image to obtain (G,B)i '. Then step 5 combines the RGB coordinates into an anaglyph (R,G.B)A-
This first method may be written as a transformation of the first image as follows:
GP (R,G,B), -* (r,g,b,W), ■* (g,b,W),' -* (G,B)A
and the transformation of the second image as follows:
Gp GB
(R5G3B)2 -^ (r,g,b,W)2 -» (g,b,W)2' -* (r,W)2' -» (R)A
where Gp is a projection transformation and GB is a balanced-brightness transformation.
In this method, due to the non-linearities of a display media, the optimal values of oci and α2 for an image may depend on the distribution of colors in the image. However even non-optimal oci and α2 may sometimes be used to reduce retinal rivalry to low levels. As the non-linearities of a display media are better accounted for in methods below, the optimal values of αi and α2 for a particular method may depend less on the distribution of colors in the image and the retinal rivalry in the image may approach zero.
Balanced Brightness Method 2
In method 2 of the present invention to create balanced-brightness anaglyphs, the non-linearities in the display device may be taken into account to some extent. First, six color hues red r, yellow y, green g, cyan c, blue b, and magenta m may be defined as follows:
y = min(R,G) - W, c = min(G,B) - W, m = min(B,R) -W, r = R-y-m -W, g = G-y-c-W, b = B-c-m-W, W = min(R,G,B).
The R,G and B coordinates may be obtained from these coordinates as follows:
R = W + y + m + r;
G = W + y + c + g; B = W + c + m + b.
The balanced-brightness condition may be summarized as follows:
y'=ocig greenish-yellow, m'=α2b bluish-magenta,
RA = W + c + y' + m', GA= W + c + g + y\ BA = W + m' + b + c
where oci and α2 may be measurable parameters which determine two hues and where RA, GA, BA are the primary coordinates of the anaglyph image. By adding the c hue to the RA
coordinate, the cyan hue CA may be absent in the anaglyph image CA=O. Also by not adding the r hue to the RA coordinate, the red hue rA may be absent in the anaglyph image, rA=0.
As in method 1 above, different display media may have different optimal values of αi and α2 which may or may not correspond to different observed hues. For one particular CRT computer monitor αi was about 9 and α2 was about 2. For one particular inkjet printer αi was about 9 and α2 was about 0.3. αi and α2 may be expected to vary greatly for display media which do not conform to certain color standards.
This second method may be implemented by first projecting the (r,y,g,c,b,m,W) color coordinates into new (g,c,b,W)' coordinates, then calculating the new (y,m)' color coordinates from the (g,c,b,W)' coordinates, and finally creating the anaglyph (R,G,B)A primary coordinates:
Red/Cyan anaglyph
Step l : (R,G,B)i -» (r,y,g,c,b,m,W)i5
Step 2: (r,y,g,c,b,m,W)i -* (g,c,bW)i\
Step 3: (g,c,b,W)j' -> (y,m)j' where y=α,g , m=α2b,
Step 4: (y,g,c,b,m,W)i' -> (R,G,B)ι' where R = W + c + y + m,
Step 5: (R)2',(G,B)i' "* (R,G,B)A
where i may be the first or second image.
Step 2 may be implemented by a method which projects the information in the initial r, y, and m coordinates into the (g',b',W) coordinates such as:
b' = βo b + β, r + β2 m , c' = β3 c, g' = β4 g + β5 y + β6 r, and
W' = β7 W + p8 r + β9 y + βiO c + β, , m .
where the βj may be selectable parameters. Many other projection methods are possible and are within the scope of the present invention. Herein, the βj parameters are called projection data. The projection data may be selected by a user according to a user's preference. Projection data may be selected which substantially transforms the relative brightness of hues in the initial image into the projected image.
This second method may be written as a transform of the first image as follows:
GP (R,G,B)i -> (r,y,g,c,b,m,W), -> (g,c,bW),' -» (y,g,c,b,m,W),' ■* (G,B)A
and the transformation of the second image as follows:
Gp GB GB
(R5G5B)2 "> (r,y,g,c,b,m,W)2 ■* (g,c,bW)2' -* (y,c,m,W)2' -» (R)A
where Gp is a projection transformation and GB is a balanced-brightness transformation.
This second method takes into account the non-linearities of a display to some degree by allowing the six hues to be independently projected into the (g,c,bW)i' coordinates.
This second method may also be applied to blue/yellow anaglyphs. For blue/yellow anaglyphs the balanced-brightness condition may be summarized as follows:
c'=αig greenish-cyan, m'=α2r bluish-magenta, RA = W + r + y + m\
GA= W + c' + g + y, BA = W + m' + y + c',
where oci and α2 may be measurable parameters and where RA, GA, BA are the primary coordinates of the anaglyph image. By adding the y' hue to the BA coordinate, the yellow hue yA may be absent in the anaglyph image, yA =0. Also by not adding the b hue to the BA coordinate, the blue hue bA may be absent in the anaglyph image, bA=0.
This second method to create balanced brightness anaglyphs may also be applied to green/magenta anaglyphs. For a green/magenta anaglyph the balanced-brightness condition may be summarized as follows:
c'=α|b greenish-cyan, m'=α2r bluish-magenta,
RA = W + r + y + m', GA= W + c' + m' + y, BA = W + m + b + c'
where otj and α2 may be measurable parameters and where RA, GA, BA are the primary coordinates of the anaglyph image. By adding the m' hue to the GA coordinate, the magenta hue ΠIA may be absent in the anaglyph image, mA=0. Also by not adding the g hue to the GA coordinate, the green hue gA may be absent in the anaglyph image, gA =0.
This second method may result in much less retinal rivalry than generally observed in color anaglyphs of the prior art. However, some retinal rivalry may still be present due to non-linearities of the display media. The non-linearities may be further taken into account by generalizing the second method. For a red/cyan anaglyph the two balanced hues relationships may be generalized as follows:
y' = Hy(g) greenish-yellow, and m' = Hm(b) bluish-magenta,
where Hy and Hm may be non-linear functions.
Balanced Brightness Method 3
In method 3 of the present invention to create balanced-brightness anaglyphs, it may be supposed that there is a linear color space RLGLBL in which the contributions of each primary color to the brightness of the first and second images in an anaglyph may be added together. First RLGLBL may be defined as follows:
R L = HR( R),
G L = HG( G), B L = HB( B),
where HR, HG, and HB may be non-linearity profiles which transform the color coordinates into linear color coordinates, and which depend on the properties of the display device.
Then the linear brightness Y L B of the first and second images may be written as
Y L B, = αGG L + αBB L,
Y LB2 = OCRR L, or
Y LB I = αGHG( G) + αBHB( B),
Y LB2 = CXRHR( R),
where Y L B I and Y L B2 represent the linear brightness of the first and second image respectively; and where αG, αo, and αc, may be measurable parameters. The balanced
brightness condition may be approximated by the relationship Y B I ~ Y B2- Then R may be written as a function of G and B as follows:
R = H2-1Ca1H0(G) H- U2HB(B)),
where H2 is a function and where αi < 1 and α2 < 1 may be measurable constants.
This third method may be implemented by first projecting the (R,G,B) primary coordinates into new (G5B)' primary coordinates, then calculating the new (R )' primary coordinate from the (G3B)' coordinates, and finally creating the anaglyph (R,G,B)A primary coordinates:
Red/Cyan anaglyph
Step l : (R5G5B)1 -* (G5B)1' ,
Step 2: (G5B)2' -* (R)2' where R = H2-^a1H0(G) + a2HB(B)) ,
Step 3: (R)2',(G,B)ι ' -» (R5G,B)A
where i may be the first or second image.
Step 1 may be implemented by a method which projects some of the infoπnation in the ( R) coordinate into the (G5B) coordinates. Methods 1 and 2 gave examples of projections which attempt to preserve some of the information in the R coordinate in the anaglyph. Either the projections of methods 1 or 2 may be used with projection data or another projection method may be used with projection data. All projection methods are intended to be within the scope of the present invention. Projection data may be selected which substantially transforms the relative brightness of hues in the initial image into the projected image.
This third method may be generally very successful in creating three-color anaglyph images with low levels of retinal rivalry. Functions H, may often be approximated by "gamma" functions as follows:
H1 (P1) = P,γi
where P1 may be R,G, and B primary colors and the γ, may be measurable parameters and are often between about 2.0 and 3.0 . Generally the optimal values of αι and α2 may create little observable retinal rivalry for all images with any distribution of colors.
This third method may be written as a transform of the first image as follows:
GP (R,G,B), -> (G5B)1' -» (G,B)A
and the transformation of the second image as follows:
Gp GB
(R,G,B)2 "> (G5B)2' -> (R)A
where Gp is a projection transformation and GB is a balanced-brightness transformation.
The three steps above are a third method of creating balanced-brightness, red/cyan anaglyphs. The first step projects the image coordinates into two primary color coordinates. The second step balances the brightness of the first and second images. The third step creates the anaglyph from the two images of the stereo pair. Similar methods of the present invention for creating balanced-brightness, yellow/blue and green/magenta anaglyphs are as follows:
Yellow/Blue anaglyph
Step l : (R,G,B)i "> (R9G)1' ,
Step 2: (R5G)2' -> (B)2' where B = H2-VIHR(R) + α2HG(G)) ,
Step 3: (B)2' ,(R5G) , ' ^ (R,G,B)A
and
Green/Magenta anaglyph
Step l : (R,G,B), -» (R9B)1' ,
Step 2: (R5B)2' -» (G)2' where G = H2 " VIHR(R) + α2HB(B)) ,
Step 3: (G)2 5XR15B1),' -* (R,G,B)A .
Balanced Brightness Method 4
In method 4 of the present invention to create balanced-brightness anaglyphs, the method 3 above for creating three-color anaglyphs in {R,G,B} primary colors is generalized to creating anaglyphs in (P15P25Qi }A primary colors. A general method for a three-color anaglyph may be written as follows:
Three-color anaglyph
Step l : (P15P25P3), -» (P15P2),' ,
Step 2: (P,,P
2)
2' -» (Q,)
2' where Q
1 =
+ (X
2Hp
2(P
2)) ,
Step 3: (Q,)2\(Pl 5P2),' -» (PI 5P25QOA.
where i is the first or second image and the balance brightness condition is:
Qi = H2-V1Hp1(PO -H a2Hp2(P2))
where Hp1, Hp2, and H2 may be non-linearity profiles which transform the color coordinates into linear color coordinates, and which depend on the properties of the display device.
The balanced brightness condition above may be a sufficient description of three- color anaglyphs with balanced brightness if it is understood that Qi is a primary color of the second image and {Pi,..,Pm} are primary colors the first image, and the condition applies to like subject matter in the first and second images.
This fourth method may be generally very successful in creating three-color anaglyph images with low levels of retinal rivalry. Functions H, may often be approximated by "gamma" functions as follows:
H1 (P1) = P,γi
where the γ, may be measurable parameters and are often between about 2.0 and 3.0 . Generally the optimal values of αi and α2 may result in little observable retinal rivalry for all images with any distribution of colors.
This fourth method of creating balanced-brightness three-color anaglyphs may be written as follows:
This fourth method may be written as a transform of the first image as follows:
GP (P15P25P3), -* (P25P3V ~> (P2,P3)A ,
and the transformation of the second image as follows:
Gp GB
(PI,P2,P3)2 -> (P23P3V -> (QOA ,
where Gp is a projection transformation, GB is a balanced-brightness transformation, and (PI,P2,QI)A are three primary color coordinates of the anaglyph. Methods 1-3 are special cases of this fourth method with different Gp and GB transformations. The projection transformation may include projection data which may be selected by a user. Projection data may be selected which substantially transforms the relative brightness of hues in the initial image into the projected image.
Balanced Brightness Method 5
The optimal values of parameters oci and α2 in methods 1-4 of the present invention may depend on the areal size of a particular hue in an image. For example a small patch of cyan surrounded by black may appear less bright than a large patch of cyan surrounded by black. Therefore, methods of adding the effects of patch size may be used to produced anaglyphs with even more accurately balanced brightness.
In method 5 of the present invention to create balanced-brightness anaglyphs, a mask image may be used to account for the variations of brightness with respect to the size of a patch of color. For simplicity of discussion, it may be assumed that the "true" brightness of a color is its brightness observed in a large patch of the color filling a substantial portion of a user's field of view. As the size of a patch of color is reduced, the apparent brightness may be reduced (or masked) by neighboring patches of color which may include a black color.
Each primary color may be masked by the presence of other primary colors nearby in the image. A masked primary color may be defined for each primary color. The R,G, and B color coordinates may be transformed into the masked color coordinates RM, GM, and BM by a masking transformation GM- For example:
RM = MR(R), GM = M0(G), and
BM = MB(B)
where MR , MG , and MB may be masking functions. For simplicity, the masking functions may be approximated as factors as follows:
RM = MRR, GM = MGG, and BM = MBB
where MR , Mc , and MB are factors. The RM, GM, and BM coordinates may be used in a balanced brightness condition. For example, the balanced brightness condition of method 3 described above gives:
RM = H2-1Ca1H0(GM) + α2HB(BM)), or
R = H2-V1HG(M0G) + α2HB(MBB))/MR,
where R, G and B may be coordinates of the same image.
The MR, MG, and MB factors may be calculated for each point i in an image by assuming a certain masking interaction TAB between the primary colors where different primary colors reduce each other's brightness and same primary color enhances its own brightness. For example, for the G an B primary colors in a red/cyan anaglyph,
M0, = ∑ TGGI1GJ - Σ (ToBjiBj), and MBl = ∑ TBBljBj - Σ (TBGJ1GJ)
while for the R primary color,
MR1 = Σ TRRlJRj
where the MR1 , MQ, . and MB, are the masking factors at each point i ; the 1\BIJ represent the interaction of the A primary color at point j in the image on the B primary color at point i in the image; and the sum is over all points j in the image. The factors TABII may be determined to decrease with increasing distance between points i and j in the image making the interaction a local interaction.
The factors MR, , MQ, , and MB, may form a brightness mask image M with three channels (MR.MG,MB). The mask image M only needs to be determined for the second image in order to implement the equal brightness condition. The image M may have less sharp edges than the second image and may be determined at lower resolution than the second image.
In some implementations of the present method, the mask image of the Qi primary color may be approximated as a constant due to the absence of other primary colors in the second image. For example, in red/cyan anaglyphs, MR=I may be assumed.
The method of calculating a mask image described above is illustrative in nature. Any method of calculating an mask image which is descriptive of the dependence of the brightness of a color on the size of the color patch is within the scope of the present invention.
This fifth method may be written as a transform of the first image as follows:
GP
(P,,P2,P3), -> (P2,P3)Γ -» (P2,P3)A ,
and the transformation of the second image as follows:
(Pi,P
2,P
3)
2 -> (P
25P
3)I' -» (P
2M
5P
3M)
2' -> (QiM)
2' -» (Q0A ,
where Gp is a projection transformation, GB is a balanced-brightness transformation, GM is a masking transformation, and (Pi, P2-QOA are three primary color coordinates of the anaglyph. Herein the mask image {MI ,M2,MQ} is called masking data M. In general, the masking data M may contain parameter values for masking functions Mi, M2, and MQ where the masking transformation GM comprises the masking functions Mg and MQ. Herein GM may be conceptually considered as a part of the Gp transformation and GM " ' may be considered as part of the GB transformation. Herein GM may be conceptually considered as a part of the Gp transformation and GM "1 may be considered as part of the GB transformation.
The present method of using a brightness mask image has been described for red/cyan anaglyphs. Similar methods of defining and using brightness mask images may be used for blue/yellow and green/magenta anaglyphs. These additional methods are within the scope of the present invention
Balanced Brightness Method 6
In a method 6 of the present invention to create balanced-brightness anaglyphs, the method 4 above is generalized to four-color anaglyphs.
If the first image is displayed in m primary colors {Pi,...,Pm} and the second image is displayed in a Qi primary color, the brightness of the first and second images, YBI and YB2 respectively, may be approximated as follows:
YB, = ∑m a; Hi(P1 ) , and YB2 = HQ, (QO
where H, and HQJ may be non-linearity profiles which transform the color coordinates into linear color coordinates; the α, may be measurable parameters; and the sum is over the m primary colors. If the balanced-brightness condition is given by YB I ~ YB2, then the balanced brightness condition may be written as:
Q1 = HQ-1CI111 α, H1(P1 ))
where HQ and the H
1 depend on the properties of the display. H
Q and the H
1 may often be approximated by "gamma" functions as follows:
V
where the γ, and γ may be measurable parameters and are often between about 2.0 and 3.0.
Four-color anaglyphs with balanced brightness may be created using the following steps:
Four-color anaglyph
Step 1 : (P,,..,Pm)2 -> (Qi)2 , where Q1 = HQ-'(Σm α, H1(P1 )) ,
Step 2: (P,,..,Pm), (Q,)2 -> (P . ,..,P011QOA,
where generally the α, may be measurable parameters. In step 1 , the second image coordinates may be used to calculate Q1. In step 2, the {Pi,..,Pm} coordinates of the first image may be combined with the Qi coordinate of the second image to create an anaglyph.
The method of creating four-color anaglyphs with balanced brightness is shown above with fewer steps than the method 4 for three-color anaglyphs because the initial images generally do not need to be projected into an intermediate image with fewer
primary colors. However a projection transformation may be used to make color modifications to the initial images. A more general method may be described as follows:
Four-color anaglyph
Step l : (P,,...,Pm), -» (Pi,..., Pn,).' ,
Step 2: (P,,...,Pm)2' "> (Qi)2' where Q1 = HQ-'(Σm α, H1(P1 )) ,
Step 3: (P,,..,Pm),' (Q1J2' -» (P,,..,Pm,Q,)A.
The balanced brightness condition Qi = HQ " '(∑m α, H,(P, )) may be a sufficient description of four-color anaglyphs with balanced brightness if it is understood that Qi is a primary color of the second image and {Pi,..,Pm} are primary colors the first image, and the condition applies to like subject matter in the first and second images.
This sixth method of creating balanced-brightness four-color anaglyphs may be written as follows:
GP First image: (P,,...,Pm), -» (P1,...,Pm),' -» (P,,...,Pm)A
Gp GB
Second image: (P,,...,Pm)2 ■* (Pι,...,Pm)2' -» (Q,)A ,
where Gp is a projection transformation, and GB is a balanced-brightness transformation, and (P i , ... ,Pm,Q i )A are four or more primary color coordinates of the anaglyph. The projection transformation may include projection data which may be selected by a user. Projection data may be selected which substantially transforms the relative brightness of hues in the initial image into the projected image.
Balanced Brightness Method 7
The optimal values of parameters oci and α2 in method 6 of the present invention may depend on the areal size of a particular hue in an image. Therefore the patch size may be taken into account using a brightness mask image as in method 5 described above.
In method 7 of the present invention to create balanced-brightness, four-color anaglyphs, a mask image may be used to account for the variations of brightness with respect to the size of a patch of color.
If the first image is displayed in m primary colors {Pi,...,Pm} and the second image is displayed in a Qi primary color, masked primary colors may be defined for each primary color Pg and Qi as follows:
where Mk and MQ may be masking functions. For simplicity, the masking functions may be approximated as factors as follows:
where Mg and MQ may be factors. The PgM and QI M coordinates may be used in an equal brightness condition. For example, the balanced brightness condition of method 6 described above gives:
QI M - = L HJQ-"I (∑m αg Hg(PgM )), or
Q, -= oH0 --i'(Σm αg Hg( MgPg ))/MQ .
where Oi and the Pg may be coordinates of the same image and the sum is over the m primary colors.
The Mg and MQ factors may be calculated for each point in an image by assuming a certain masking interaction TAB between the primary colors. For example, for the {Pi,...,Pm} primary colors
Mg1 = ∑j Tgg,jPgj - ∑j∑h TghijPhj ,
where the Mg, may be the masking factors at each point i; the Tgh,j represent the interaction of the Ph primary color at point j in the image on the Pg primary color at point i in the image; and the sum is over all points j in the image and primary colors Ph where h is not equal to g. The factors Tghυ may be determined to decrease with increasing distance between points i and j in the image making the interaction a local interaction. The factors Mg, may form a brightness mask image M with channels {Mi,...,Mm,Mo} or mask images {Mi,...,Mm} and {MQ} . The mask image M only needs to be determined for the second image in order to implement the equal brightness condition. The image M will generally may have less sharp edges than the second image and may be determined at lower resolution than the second image.
The mask image of the Qi primary color may be approximated as follows:
MQ1 = Σj TQQyQlj
where TQQ,J is the interaction between brightness of the Qi primary color at point j with the brightness of the Qi primary color at point i. The mask image of the Qi primary color may be approximated as a constant due to the absence of other primary colors in the second image. In other words, MQ=I may be assumed. Then
Q, - = HLJQ--I'(Σ αg Hg( MgPg )).
The method of calculating a mask image described above is illustrative in nature. Any method of calculating a mask image which is descriptive of the dependence of the brightness of a color on the size of the color patch and may be determined for the second image rendered in (P1, ...Pm} coordinates or in (Qi } coordinates is within the scope of the present invention.
This seventh method may be written as a transform of the first image as follows:
GP
(p,,...,pm)i -» (P,,...pmy ■* (P. ,...pm)A ,
and the transformation of the second image as follows:
(Pl , ... ,Pm)2 ■* (P1 ,- - .Pm)2' ^ (Pl M,... PmM)2' "* (QI M)2' "> (QOA ,
where Gp is a projection transformation, GB is a balanced-brightness transformation, GM is a masking transformation, and (PI,P2,QI)A are three primary color coordinates of the anaglyph. Herein the mask image (Mi, ... ,M1115MQJ is called masking data M. In general, the masking data M may contain parameter values for masking functions Mg and MQ where the masking transformation GM comprises the masking functions Mg and MQ. Herein GM may be conceptually considered as a part of the Gp transformation and GM "1 may be considered as part of the GB transformation.
Creating True-Depth Anaglyphs
Generally, anaglyphs of the present invention with balanced brightness may be created in two stages. The first stage concerns determining the parameters of a balanced- brightness transformation GB- Herein these parameters are called brightness
transformation data or GB data. The second stage concerns using the GB data to create balanced-brightness anaglyphs from stereoscopic image pairs. In theory, the GB data only needs to be determined once for a display apparatus. The GB data may be stored and used repeatedly to create balanced-brightness anaglyphs. However, a range of GB data may be selectable by the user to suit a user's preferences. The relative brightness of the first and second images may be selectable by a user. The first image may be selectable as the left or right eye image by a user. Furthermore the projection transformation Gp may contain a set of parameters which may affect the color distribution in the displayed anaglyphs. Herein these parameters are called Gp data. Generally, the Gp data may be selectable by a user. Herein the Gp data may include: the relative brightness of the first and second images; which of the left or right images is the first image of the anaglyph; and the method of projecting the initial primary color into the primary colors of the first image of an anaglyph.
Generally software may be used to help determine the GB data and to create balanced-brightness anaglyphs. A software program may provide tools which help determine the GB data for a certain display. A software program may provide tools which help create anaglyphs using the GB data. A software program may provide tools which allow the Gp data to be selected by a user. A display apparatus may comprise GB data in the apparatus which may be used to create balanced brightness anaglyphs without requiring the user to determine the GB data. A display apparatus may comprise transformations Gp and/or GB for creating anaglyphs with balanced brightness from full- color stereoscopic images.
Figure 9 depicts a flowchart of the process of determining the GB data for three- color anaglyphs. The display apparatus provides two primary colors {Pi,P2} for displaying the first image of an anaglyph and a primary color (Qj } for displaying the second image of an anaglyph; and includes: a first viewing filter for viewing the {Pi, P2) primary colors; and a second viewing filter for viewing the (Qi } primary color. The process comprises the following steps: (1) reading the coordinate values of a test image
902; (2) projecting the image coordinate values into the {Pi,P2} primary colors using a Gp transformation 904, 906; (3) selecting initial GB data 908; (4) calculating the (Q1) coordinate values using the GB data and GB transformation 908, 910, 912; (5) creating an anaglyph from the {Pi,P2} and {Qi } values 914; (6) displaying the anaglyph 916; (7) viewing the anaglyph through the first and second filters 918; and (8) if the anaglyph has balanced brightness contrast the GB data may be stored 922 and used to create anaglyphs 924; if the anaglyph does not have balanced brightness contrast, the GB data may be adjusted 920 and new {Qi } values may be calculated using the new GB data 920, 910, 912.
Figure 10 depicts a flowchart of the process of creating and displaying three-color, balanced-brightness anaglyphs using stored GB data. The flowchart depicts the first and second images being processed in parallel, however they may also be processed sequentially. The process comprises the following steps: (1) reading the first image coordinate values 1002; (2) projecting the first image coordinate values into the {Pi,P2} i primary colors using a Gp transformation 1004, 1006; (3) reading the second image coordinate values 1008; (4) projecting the second image coordinate values into the {PiΛh primary colors using a Gp transformation 1010, 1012; (5) reading the GB data 1014; (6) calculating the {Qi }2 coordinate values using the GB data and the GB transformation 1014, 1016, 1018; (7) creating an anaglyph from the {Pi,P2} ι and (Qi }2 values 1020; (8) displaying the anaglyph if desired 1022, 1024. The anaglyph {P|,P2,Qi } may also be stored in a digital file. A display apparatus performing this process may usually display the anaglyphs. A software program performing this process may usually store or display the anaglyphs.
Figure 1 1 depicts a flowchart of the process of determining the GB data for four- color anaglyphs. The display apparatus provides three primary colors (Pi5P2, P3) for displaying the first image of an anaglyph and a primary color {Qi } for displaying the second image of an anaglyph; and includes: a first viewing filter for viewing the {Pi,P2,P3} primary colors; and a second viewing filter for viewing the {Qi } primary
color. The process comprises the following steps: (1) reading the coordinate values of a test image 1102; (2) projecting the image coordinate values into the (Pi3P23P3) primary colors using a Gp transformation 1104, 1106; (3) selecting initial GB data 1108; (4) calculating the (Qi } coordinate values using the GB data and GB transformation 1108, 1110, 1112; (5) creating an anaglyph from the (Pi3P23P3) and (Qi ) values 1114; (6) displaying the anaglyph 1116; (7) viewing the anaglyph through the first and second filters 1118; and (8) if the anaglyph has balanced brightness contrast the GB data may be stored 1122 and used to create anaglyphs 1124; if the anaglyph does not have balanced brightness contrast, the GB data may be adjusted 1120 and new (Qi ) values may be calculated using the new GB data 1120, 1110, 1112.
Figure 12 depicts a flowchart of the process of creating and displaying four-color, balanced-brightness anaglyphs using stored GB data. The flowchart depicts the first and second images being processed in parallel, however they may also be processed sequentially. The process comprises the following steps: (1) reading the first image coordinate values 1202; (2) projecting the first image coordinate values into the (Pi,P2,P3) i primary colors using a Gp transfoπnation 1204, 1206; (3) reading the second image coordinate values 1208; (4) projecting the second image coordinate values into the (Pi,P2,P2}2 primary colors using a Gp transformation 1210, 1212; (5) reading the GB data 1214; (6) calculating the (Qi )2 coordinate values using the GB data and the GB transformation 1214, 1216, 1218; (7) creating an anaglyph from the (Pi5P23P3) i and (Qi)2 values 1220; (8) displaying the anaglyph if desired 1222, 1224. The anaglyph (P i,P2, P3, Qi ) may also be stored in a digital file. A display apparatus performing this process may usually display the anaglyphs. A software program performing this process may usually store or display the anaglyphs.
A digital file may be used to store the anaglyphs of the present invention with four or more color values (P|,...Pm,Qi } as an array of m+1 color coordinates representing the color coordinates of each pixel point j in a digital image (Pi,...Pm,Qi },. Alternatively the anaglyph may be stored in m+1 arrays of coordinates { P i } j, ... , ( Pm } j. { Q i } j. Alternatively
the anaglyph may be stored in two arrays of coordinates {Pij,...,Pm}j,{Qi }j- The digital file of the present invention may be compressed by methods of the prior art.
Control of Retinal Rivalry
In some embodiments of the present invention, the difference in brightness contrast between the first and second images may be used to define a measure of retinal rivalry in an anaglyph image. A measure of retinal rivalry may allow the retinal rivalry to be controlled across the entire anaglyph image for all colors. A certain low amount of retinal rivalry may be used to increase the color gamut of some anaglyphs. In contrast in the prior art, the reduction of retinal rivalry over some regions of an anaglyph may create high levels of retinal rivalry in other regions of the anaglyph.
The methods of creating balanced brightness anaglyphs described herein provide a method of determining the values of Qi in the anaglyph. The resulting Qi values may be approximations of the ideal values Qi' that truly balance the brightness contrast in the anaglyph. The difference in the calculated values Qi and the ideal values Qi' may be used to create a measure of the retinal rivalry Z' of the stereo view or anaglyph:
Z' = |Q, - Q, '|/Q,
where the (Qi - Qi') factor is proportional to the retinal rivalry and the 1 /Qi factor is a normalization factor. Calculating Q' may require an exact model of the brightness of a color. However, Qf may also be approximated by models which determine the values of QI B which minimize retinal rivalry.
One embodiment of the present invention provides a method to measure the retinal rivalry Z in an anaglyph using the difference Qi - Q
I B in the actual Qi values in an anaglyph from the values Q
I B calculated by a particular method to minimize the retinal rivalry in an image:
.
where λ is a scaling ratio which may take into account a different scaling between Qi and Qi
B- The (Qi
B - λQi) factor may be proportional to the retinal rivalry. The 1/Qi factor is a normalization factor. In general, λ may be selected to minimize the average Zχ over an anaglyph image. Or if the scaling ratio is known, λ may be set to the scaling ratio value. Then the minimum of Zχ may be a measure of the retinal rivalry,
Z = MINλ(|Q,B - λQ,|/Q1 B).
Once λ is determined, Z may be written without loss of generality as:
B.
Z is a color coordinate unique to stereoscopic images which may be used to measure the retinal rivalry of like subject matter in an anaglyph image. Herein, Z is called the retinal rivalry coordinate.
The retinal rivalry coordinate Z may be used to compare various methods of creating anaglyphs including methods of the prior art. In order to determine Z, the balanced brightness values QI B may be determined by a selected method. The more accurately QI B is determined, the more accurately Z measures the retinal rivalry. If the {Pι,...,Pm}2 coordinates of the second image are known, Z may be determined in a straight forward way. The {Pi,...,Pm}2 coordinates of the second image may be used to calculate {QI B}2 for the second image. Then Z may be determined by comparing the values of {QI B}2 with the values of {Qi }2 in the anaglyph. If the {Pi,...,Pm}2 coordinates of the second image are not known, {QI B} i may be calculated using the values of {Pi,...,P111 } i from the first image for like subject matter in the second image. Then Z may
be determined from the values of {QI B} i for the first image and the values of {Qi }2 in the second image.
The retinal rivalry coordinate Z of the present invention may be used to: (1) determine an average retinal rivalry in an anaglyph; (2) determine the maximum values of retinal rivalry in an anaglyph; and (3) determine a distribution of retinal rivalry in an anaglyph image. These tools may be provided in a software program.
Another measure of the retinal rivalry is:
Z5 = MINλ((Q1 B - λQ,)/Q,B )
where λ may be chosen to minimize the average magnitude of Z
s over an image. Once λ is determined, Z
s may be written without loss of generality as:
•
Zs contains the sign of the retinal rivalry coordinate, Z = |ZS|. In general a measure Z of retinal rivalry can be any function of Zs
Z = G(Z5).
Some embodiments of the present invention provide methods to increase the color gamut observed in an anaglyph image by using selected amounts of retinal rivalry. One of the problems with anaglyphs of the prior art is that they may contain retinal rivalry that is uneven over the image, and they often contain high levels of retinal rivalry for some hues. In one embodiment of the present invention, the Z
s coordinate may be used to create anaglyph images with a controlled amount of retinal rivalry. First a method may be used to determine a balanced brightness Qm coordinate. Then the values of Z
s may be selected over the image. Herein, the selected values of Z
5 are called the rivalry data or Z
s
data. The values of Q
I B determined by a brightness transformation G
B may be transformed into the Qi values of an anaglyph with a selected amount of retinal rivalry using a rivalry transformation Gz. Gz may be written as follows:
where λ =1 since Qi and QI B may have the same scale. The rivalry data Z8 may be a user selectable constant or a user selectable function of any color property or coordinate or region of the image. A positive Z5 values may generally enhance the observed saturation of hues in the first image. A large negative Z5 values may generally add the hues associated with the Qi primary color to the observed color gamut in the stereo view. The Z5 data may be varied selectively in order to add hues to various areas of an anaglyph image.
A display apparatus may comprise rivalry data Z5. The rivalry data Z5 in a display apparatus may be selected from a range by the user according to a user's preference. Similarly, a software program may comprise data Z5. The rivalry data Z5 in a software program may be selected from a range by the user according to a user's preference.
One embodiment of the present invention includes methods for creating anaglyphs with a selected amount of retinal rivalry using a rivalry transformation Gz comprising rivalry data Z5. This method may be written as a transform of the first image as follows:
GP (P,,...,Ps)1 -* (P,,...Pm), ' -> (P,,...Pm)A ,
and the transformation of the second image as follows:
(P,,...,Ps)2 ^ (Pl,...Pm)2' -» (QlB)2' "* (Ql)A ,
where s may be different from m, Gp is a projection transformation, GB is a balanced- brightness transformation, GM is a masking transformation, and (Pi,...Pm,Qi)A are m+1 primary color coordinates of the anaglyph. Herein Gz may be conceptually considered as a part of the GB transformation and GM '•
Figure 13 depicts a flowchart of the process of creating and displaying four-color, anaglyphs with a selected amount of retinal rivalry using stored Z5 data. The process includes some of the steps shown in figure 12 and not shown in figure 13. The process comprises the following steps: (1) reading the first image coordinate values 1202; (2) projecting the first image coordinate values into the {Pi,P2,P3} i primary colors using a Gp transformation 1204, 1206; (3) reading the second image coordinate values 1208; (4) projecting the second image coordinate values into the {Pi,P2,P2}2 primary colors using a Gp transformation 1210, 1212; (5) reading the GB data 1214; (6) calculating the (Qiβh coordinate values using the GB data and the GB transformation 1214, 1216, 1302; (7) reading the retinal rivalry data 1304; (8) determining the (Qi }2 using the retinal rivalry transformation Gz 1306; (9) creating an anaglyph from the {Pi,P2,P3} i and (Qi }2 values 1220; (8) displaying the anaglyph if desired 1222, 1224. The anaglyph (Pi,P2;P3,Qi } may also be stored in a digital file. A display apparatus performing this process may usually display the anaglyphs. A software program performing this process may usually store or display the anaglyphs.
The methods of the present invention to control retinal rivalry may be used to broaden the color gamuts of three-color anaglyphs. Using positive Zs near the white point may enhance the hues which may be de-saturated by the shift in the white point toward the hue. For example, in a red/cyan anaglyph, using positive Z5 for the cyan and green hues may cause the appearance of the cyan hue and an increased saturation of the green hues near cyan. Using negative Zs near the blue or yellow hues may shift these hues toward purple and orange hues respectively.
Figure 14 depicts the color gamut 1406 of a red/cyan anaglyph with a selected range of retinal rivalry Z5. The color gamut for the case of no retinal rivalry Z8=O is shown as a one-dimensional line segment 1402 joining the Pi and P2 primary colors. A upper limit of the selected Z5 (positive Z5) is shown as a curve 1404 on the green and cyan side of the no-rivalry line segment 1404. The positive values of Zs may be used to increase the saturation of the green and cyan hues in the color gamut due the increased distance from the apparent white point Wp' viewed through the first viewing filter. The lower limit of the selected Z5 (negative Z5) is shown as a curve 1404 on the red side of the no-rivalry line segment 1404. The negative values or Z5 may be used to improve the yellow hues and add purple hues to the color gamut of the anaglyph. Figure 14 depicts the combined effects of using both positive and negative Z5 to expand the color gamut of anaglyph images. Similarly, selected amounts of retinal rivalry may be used to expand the color gamut observed in blue/yellow and green/magenta anaglyphs and in four-color anaglyphs.
One aspect of the present invention is a software program which provides methods to measure the retinal rivalry in anaglyph image. One aspect of the present invention is a software program which provides tools to create anaglyphs with selected amounts of retinal rivalry. One aspect of the present invention is a software program which provides tools for measuring the retinal rivalry of existing anaglyphs. One aspect of the present invention is a display apparatus which comprises Z data. One aspect of the present invention is a display apparatus which allows the Z data to selected from a range by the user.
Subtractive Primary Colors
In applying the balanced brightness condition to printed anaglyph images, the brightness property may be determined for subtractive primary color values. The dependence of the brightness on subtractive primary colors may be determined by converting the subtractive primary values {mPi,...,mPs} into additive primary color
values {Pi,...,Pm} . The number of subtractive primary colors may be different from the number of additive colors. If the images are originally in additive color coordinates, the additive color coordinate values may be transformed into the gamut of the subtractive primary colors. Then the additive primary colors may be used to determine a Qi primary color value. Then the Qi primary color values may be transformed into the subtractive primary color values.
As noted above, the present invention is applicable to software, display apparatus and special filters and is believed to be particularly useful for creating, displaying and viewing digital anaglyph images with wide color gamuts without retinal rivalry or with a selectable amount of retinal rivalry. The present invention should not be considered limited to the particular examples described above, but rather should be understood to cover all aspects of the invention as fairly set out in the attached claims. Various modifications, equivalent processes, as well as numerous structures to which the present invention may be applicable will be readily apparent to those of skill in the art to which the present invention is directed upon review of the present specification. The claims are intended to cover such modifications and devices.