CN105741235A - Visual rivalry improving method of complementary color three-dimensional image for improving color fidelity - Google Patents

Abstract

The invention provides a visual rivalry improving method of a complementary color three-dimensional image for improving color fidelity. The method comprises the steps of transforming symmetrical left view and right view of a three-dimensional image from an RGB color space to a CIE L*a*b* color space; then calculating the sensing area of a cyan lens in the CIE L*a*b* color space; matching and mapping the right view to the two-dimensional sensing area of the cyan lens, and calculating corresponding a* and b* values of each pixel in the matched and mapped right view; adjusting the brightness of the left view and the right view based on a chromatic value, and separately calculating a corresponding L* value of each pixel in the adjusted left view and right view; inversely transforming the processed left view and right view from the CIE L*a*b* color space to the RGB color space to synthesize a red and cyan complementary color three-dimensional image. The method can improve the color fidelity of the complementary color three-dimensional image on the premise of effectively reducing the visual fatigue in view of the three-dimensional image and obtaining more comfortable three-dimensional visual experience.

Description

Improve the anaglyphs visual rivalry ameliorative way of color fidelity

Technical field

The present invention relates to a kind of anaglyphs synthetic technology, especially improve in anaglyphs visual rivalry phenomenon, improve the dark purple anaglyphs synthetic technology of color fidelity.

Background technology

Along with virtual reality technology was stepped into " the virtual reality first year " in 2016, virtual reality applications initially enters commercialization and promotes the stage.Correspondingly, the level of hardware of stereoscopic display device constantly promotes, and people are unprecedentedly surging to demand and the enthusiasm of stereo display.Stereo display technique can divide two big classes: a class is to wear the Auto-stereo display mode of any aid, i.e. bore hole formula stereo display technique, show including body, based on the holographic display etc. of the Auto-stereo display of liquid crystal display, non-spectacle, this kind of technology still locates conceptual phase at present, due to all many-sided restrictions such as technology, great number equipment cost and film source making, actual business promotion uses and also needs the long period；The another kind of stereo display technique being to need to wear auxiliary equipment, foundation binocular parallax principle, the equipment of representative is Helmet Mounted Display and anaglyph spectacles.Helmet Mounted Display is only applicable to single viewing, after stereo display hardware device obtains tremendous development, volume and weight based on the helmet of Helmet Mounted Display is substantially reduced, developing into comparatively light glasses form, also call with anaglyph spectacles now, price is still higher, it is professional anaglyph spectacles referred to here as this kind of anaglyph spectacles, such as the Rift glasses of Oculus company of the U.S., price $ 599, the PC that the performance of about 1000 dollars is powerful separately need to be joined.Another kind of anaglyph spectacles is the conventional stereo glasses only with three-dimensional display function, belong to popular consumer goods, cheap, it is consumer level anaglyph spectacles referred to here as this kind of anaglyph spectacles, the polarized stereoscopic glasses that be divided into the anaglyphic spectacles divided based on color, divides based on light and the shutter stereo glasses based on the time-division.Polarized stereoscopic glasses needs to be equipped with special stereo projecting equipment, is typically employed in the occasion such as movie theatre, exhibitions demonstration；Anaglyphic spectacles separates two eye pattern pictures by the color filtration of eyeglass, the stereo-picture issued on the internet and three-dimensional video-frequency generally use this anaglyph spectacles to watch, it is widely used in the personal consumption fields such as PC, mobile phone, iPad, there is the most vast customer group.

At present, virtual reality hardware device still suffers from many problems demand in actual applications and solves, including: (1) film source is without versatility: different stereoscopic display devices can only use the film source of oneself professional format, the not versatility of this film source is particularly evident on the impact of emerging stereoscopic display device popularization and application, these stereoscopic display devices are releasing initial stage film source seldom, and will reach certain customer volume have to continue to increase a large amount of input in the software aspects that film source produces.(2) volume and weight of Helmet Mounted Display and professional anaglyph spectacles is still excessive, and compared with the augmented reality holographic glasses Hololens of Microsoft, it appears stupid much heavier, comfort level and feeling of immersion on viewing have impact.

Make a general survey of all kinds of stereo display technique, the operation principle of its stereoscopic display device and the characteristic of stereo-picture are analyzed, it is found that the complementary stereo display technique only divided based on color possesses the basic condition solving the problems referred to above.Complementary stereo display technique is stereo display technique the earliest, at the initial stage of stereoscopic display device hardware technological development, some problems existed due to color point technology are not solved very well, and stereo display effect is not ideal enough, and gradually faded from memory, technical development is also therefore slow.But, current after fast development of stereo display hardware technology, the inherent character of complementary stereo display technique but brings opportunity for solving new problem, analyzes and can obtain as drawn a conclusion: (1) can solve film source versatility problem principle.Complementary stereo display use anaglyphs or auto-stereoscopic video can compliant applications in all stereoscopic display devices, such as: can using the image of two complementary color passages or video signal as professional anaglyph spectacles the input signal source of two LCD display, it is achieved stereo display；The input signal source of left and right projector equipment in the image of two complementary color passages or video signal can also being shown as polarized stereoscopic, it is achieved stereo display.It is to say, anaglyphs or three-dimensional video-frequency manufacturing technology possess the condition showing film source production method as a kind of General stereoscopic.The film source changing current various stereoscopic display devices is made the state being respectively battalion by unified film source manufacturing technology, same film source is avoided to repeat to make the cost waste caused, various stereoscopic display devices both can use the film source of oneself professional format, general film source can also be used, the abundant rapidly of stereo display film source will be promoted.This result, the dissemination newly going out stereoscopic display device is more great, and new equipment just can use a large amount of existing film source for Consumer's Experience at the beginning of releasing, and its popularization is had important facilitation.(2) miniaturization of stereoscopic display device it is greatly facilitated.Auto-stereoscopic imaging technique utilizes two Color Channels to realize stereoscopic vision, as long as viewing equipment has corresponding colour filter function, the accessories such as required power supply are shown without LCD, therefore in principle, current anaglyphic spectacles can be made the form of contact lens, utilizing the production technology being similar to contact lens or U.S. pupil to realize, this will bring a qualitative leap of stereoscopic display device miniaturization.(3) it is of value to the stereos copic viewing improving the handheld mobile device with huge customer group to experience.Along with the explosive growth of Helmet Mounted Display and professional anaglyph spectacles equipment, the stereo display mode of the handheld mobile device such as mobile phone, iPad also begins to be subject to the concern of some VR equipment vendors.Google company prepares the stereo display glasses that investment cellphone subscriber is new, but does not break away from the thinking of their simple solid glasses Cardboard, it is necessary to develop special mobile phone operating system and chip.According to Such analysis, for handheld mobile device, auto-stereoscopic shows the potential ability possessing the most convenient viewing mode, for the user worn glasses, complementary color electrostatic pad pasting is adsorbed on the daily eyeglass worn, the user that just can solve easily to wear glasses in other stereoscopic display devices watches the puzzlement of stereo-picture and three-dimensional video-frequency, and certainly, this kind of user can also use the form of contact lens to watch more easily future.

In view of the above-mentioned advantage that the inherent characteristics of complementary stereo display technique embodies when development of new techniques, the present invention studies the anaglyphs synthetic technology in complementary stereo display technique, it is directed to most popular dark purple anaglyphs, viewing equipment is dark purple anaglyphic spectacles, and two eyeglass is red color filter and cyan light filter respectively.

The concept related in anaglyphs synthesis includes:

(1) binocular parallax stereoscopic vision and anaglyphs

When mankind's eyes watch the object of certain distance, the visual pattern received due to the existence of two optical axis spacing, left eye and right eye is different, is a pair stereo pairs, the scene image that wherein left eye is seen is called that the scene image that left view, right eye are seen is called right view.Two width image fine distinctions can be merged by brain, perceives physiology depth cueing, thus producing third dimension.Anaglyphs synthetic technology using color solid image to as film source, left view and right view to stereo image pair carry out computing, generate a width anaglyphs, two kinds of image informations of superposition on this image: one is that left eye passes through red color filter and should observe the image information perceived, calculated by left view and obtain, it is called " perception left view ", another is that right eye passes through cyan light filter and should observe the image information perceived, calculated by right view and obtain, be called " perception right view ".When watching anaglyphs by dark purple anaglyphic spectacles, the perception view that two eyes are each seen by brain merges, thus understanding depth information, produces third dimension.

(2) visual rivalry

Visual rivalry refers to when the profile of the presented image of eyes, brightness or color differ greatly, left view and right view cannot be fused into a stereoscopic image by brain completely, cause the dynamic alternate of visually-perceptible: sometime, the image sensation that left eye is seen is clear, the visual stimulus of right eye is suppressed, and the next moment is then contrary.The visual stimulus of right and left eyes is all striving that the leading position being in vision but can not produce to merge.Visual rivalry phenomenon is the main cause affecting anaglyphs stereoscopic vision perception stability.

(3) three elements of color

The three elements of color refer to tone (or claiming form and aspect), saturation and brightness, in order to describe the human eye perception to color.Wherein, the classification of tone reflection color, depends on the dominant wavelength of light that object sends or reflects, is generally identified by color designation；Saturation refers to gradation of color, more high containing colour content, and saturation is more high, and color is more pure；Brightness refers to the relative bright-dark degree of color.

Tone in color three elements and saturation are generically and collectively referred to as colourity.

(4) color space

Color space is also referred to as color model, and the color space related in anaglyphs synthesis has: (a) RGB color: R, G, B represent Red Green Blue respectively；(b)CIEL^*a^*b^*Color space: L^*For psychology brightness, a^*、b^*It is feel consistent two chrominance channels with tone and saturation；C () CIE XYZ color space: developed by rgb space, is a kind of basic skills measuring color, X, Y, Z are the three primary colories of this color space；D () hsv color space: H is tone, S is saturation, and V is brightness.

Common anaglyphs composition algorithm is broadly divided into two big classes, is empirical composition algorithm and calculation type composition algorithm respectively.Empirical composition algorithm is left out spectrum parameter and the anaglyphic spectacles absorption curve function to different-waveband of display device, with the passage addition method for main representative.Its main thought is the Color Channel utilizing two kinds of complementary colors each other, reaches the effect of separate colors in conjunction with corresponding anaglyphic spectacles thus producing three-dimensional perception.But, the passage addition method often because serious visual rivalry, makes people do not feel like oneself, easily causes visual fatigue.Modified hydrothermal process mainly has two classes on this basis: a class is to utilize greyscale transformation, first the left view of stereo image pair or left view and right view is become gray level image simultaneously, then synthesizes by the passage addition method；Another kind of, it is utilize gamma correction, cyan component is mapped in red component, to improve the stereoscopic visual effect of the anaglyphic map of synthesis.But these innovatory algorithm are in accordance with experience to be adjusted the mapping parameters in the passage addition method, not tight logical analysis, after causing the image that scene is different is synthesized, stereoeffect difference is very big.Calculation type composition algorithm considers spectrum parameter and the anaglyphic spectacles absorption curve function to different-waveband of display device, and it represents algorithm mainly XYZ method, Lab method and HSV method.XYZ method is the stereo pairs of input to be converted into CIE XYZ color space be calculated synthesis, is minimized the color error between original image and the image that perceived by anaglyphic spectacles by the method for least square.Due to the heterogeneity of CIE XYZ color space, there is certain cross-color in the anaglyphs of synthesis, and visual rivalry phenomenon is obvious.Lab rule is that XYZ method is converted into CIEL^*a^*b^*Color space, it is intended to utilize CIEL^*a^*b^*The perception uniformity of color space obtains better stereoeffect, and this algorithm effectively improves cross-color problem, but the anaglyphs of its synthesis still suffers from obvious visual rivalry phenomenon.HSV method is to directly utilize hsv color space to more conform to the human vision characteristic to color description, carries out the coupling of saturation in hsv color space, then contravariant shifts to rgb space synthesis anaglyphs.This algorithm reduces the impact of visual rivalry to a certain extent, but its color distortion ratio is more serious, especially there will be bigger distortion at blue and red area.

Summary of the invention

In order to overcome the deficiencies in the prior art, the present invention provides a kind of anaglyphs visual rivalry ameliorative way improving color fidelity, matching condition in utilizing tone invariance to adjust as color, in order to balance the relation of visual rivalry and cross-color so that the dark purple anaglyphs of synthesis can obtain more accurate color reconstruction while reducing visual rivalry；And simultaneously by regulating the purpose reducing visual rivalry based on the luminance dynamic of tone and saturation, improve the stability of stereoscopic vision perception；Can at the visual fatigue in effectively reducing stereo-picture viewing, it is thus achieved that under the premise of more comfortable stereo vision experience, improve the color fidelity of anaglyphs.

The technical solution adopted for the present invention to solve the technical problems comprises the following steps:

First, the left view of stereo image pair and right view are converted into CIEL by RGB color respectively^*a^*b^*Color space；Then, cyan eyeglass is calculated at CIEL^*a^*b^*The sensing region of color space；Using tone invariance as constraints, right view coupling is mapped to the two-dimentional sensing region of cyan eyeglass, calculates a that after coupling maps, in right view, each pixel is corresponding^*、b^*Value；Based on colourity (tone and saturation) value, left view and right view carried out brightness regulation, and calculate the L that the pixel each with in right view of the left view after adjustment is corresponding respectively^*Value；Finally, by process after left view and right view by CIEL^*a^*b^*Color space contravariant shifts to RGB color, synthesizes dark purple anaglyphs.Mainly comprise the steps:

Step one, at RGB color, the left view pixel I to stereo image pair_L,RGBWith right view pixel I_R,RGBTri-channel value of respective R, G, B are normalized so that it is be distributed in [0,1] scope；As intermediary, left view pixel and right view pixel are converted into CIEL by RGB color respectively using CIE XYZ color space^*a^*b^*Color space,In formula, I_{L,RGB_R}、I_{L,RGB_G}、I_{L,RGB_B}Respectively left view pixel I_L,RGBCorresponding tri-channel value of R, G, B, I_{L,XYZ_X}、I_{L,XYZ_Y}、I_{L,XYZ_Z}It is then left view pixel I respectively_L,RGBAt the X of CIE XYZ color space, Y, Z component value, I_{R,RGB_R}、I_{R,RGB_G}、I_{R,RGB_B}Respectively right view pixel I_R,RGBCorresponding tri-channel value of R, G, B, I_{R,XYZ_X}、I_{R,XYZ_Y}、I_{R,XYZ_Z}It is then that right view pixel is in the X of CIE XYZ color space, Y, Z component value respectively；Cs is the transformation matrix being converted into CIE XYZ color space by RGB color,

Utilize CIE XYZ color space and CIEL^*a^*b^*The transformation relation of color space, is converted into CIEL further respectively by the left view pixel and right view pixel that transform to CIE XYZ color space^*a^*b^*Color space,In formula, I_{L,Lab_L}、I_{L,Lab_a}、I_{L,Lab_b}Respectively left view pixel I_L,RGBAt CIEL^*a^*b^*The L of color space^*、a^*、b^*Component value；I_{R,Lab_L}、I_{R,Lab_a}、I_{R,Lab_b}It is right view pixel I respectively_R,RGBAt CIEL^*a^*b^*The L of color space^*、a^*、b^*Component value, the expression formula of f () isIt is the reference white point of CIE XYZ color space, [X_whiteY_whiteZ_white]^T=Cs × [111]^T；

Each pixel in left view and right view is carried out above-mentioned computing, thus completing stereo pairs from RGB color to CIEL^*a^*b^*The conversion of color space；

Step 2, one panchromatic right view comprising all colours of definition, its color is at RGB color three-dimensional matrice C_{256×256×256}Storage, the every one-dimensional element of matrix is the integer from 0 to 255；Calculate this panchromatic right view through after cyan eyeglass at CIEL^*a^*b^*The distributed areas of color space, are cyan eyeglass at CIEL^*a^*b^*The sensing region of color space, its computational methods are as follows:

By C_{256×256×256}Replace with the V 100 width HS figure changed from 0 to 99 integer, V is the luminance component in hsv color space, HS figure is V when being certain constant in span, using tone H as transverse axis, saturation S is as the 2 d plane picture of the longitudinal axis, and to the right, span is the integer of 0 to 359 to tone H axle positive direction level, vertically downward, span is the integer of 0 to 99 to saturation S axle positive direction；If correspondence HS during V=i_iFigure, i is integer and 0≤i≤99, I_{HS_Ri}、I_{HS_Gi}、I_{HS_Bi}Represent HS respectively_iThe color of figure is at tri-channel value of R, G, B of RGB color, I_{RP,XYZ_Xi}、I_{RP,XYZ_Yi}、I_{RP,XYZ_Zi}Represent that it is at CIE XYZ color space tri-component values of corresponding X, Y, Z respectively, thenIn formula, I_{RP,Lab_Li}、I_{RP,Lab_ai}、I_{RP,Lab_bi}Represent HS respectively_iThe color of figure is at CIEL^*a^*b^*The corresponding L of color space^*、a^*、b^*Component value,Be according to cyan eyeglass, the absorption curve function of different-waveband color is calculated, by the transformation matrix of RGB color to CIEXYZ, [X_{R_white}Y_{G_white}Z_{B_white}] it is that CIE XYZ color space corresponds to transformation matrix A_RReference white point, [X_{R_white}Y_{R_white}Z_{R_white}]^T=A_R×[111]^T；

According to calculated all I_{RP,Lab_Li}、I_{RP,Lab_ai}、I_{RP,Lab_bi}, at CIEL^*a^*b^*Color space drawing 3 D graphics, obtains the three-dimensional perception region of cyan eyeglass.This sensing region be distributed as a series of point bunch body, be approximately perpendicular to a from spatially observing it^*-b^*Plane.By three-dimensional perception region projection to a^*-b^*Plane, obtains the region that namely two dimension sensing region is curve one and curve two surrounds, and the equation of curve one is b₁=-5.1146 × 10^-4·(a₁)²-0.8724·(a₁)+22.5848, the equation of curve two is b₂=0.0029 (a₂)²-0.89·(a₂)-5.5028, in formula, b₁、a₁The respectively b of curve one^*Component value and a^*Component value, b₂、a₂The respectively b of curve two^*Component value and a^*Component value；

Step 3, at a^*-b^*Plane carries out coupling and maps, and the color of right view is mapped to CIEL^*a^*b^*The two-dimentional sensing region of cyan eyeglass in color space；

The a of border, curve one upper right side exterior pixel after note coupling mapping^*、b^*Component value is I_{RM,Lab_au}、I_{RM,Lab_bu}, tone value isA before Corresponding matching mapping^*、b^*Component value I_{R,Lab_au}、I_{R,Lab_bu}, tone valueIt is derived by

The a of border, curve two lower left exterior pixel after note coupling mapping^*、b^*Component value is I_{RM,Lab_ad}、I_{RM,Lab_bd}, tone valueA before Corresponding matching mapping^*、b^*Component value I_{R,Lab_ad}、I_{R,Lab_bd}, tone valueIt is derived by

By on the two-dimentional sensing region boundary curve one of the pixel-map outside border, curve one upper right side to cyan eyeglass, have

Try to achieve I_{RM,Lab_au}、I_{RM,Lab_bu}Solution；All pixels outside for border, curve one upper right side perform this operation；

By on the two-dimentional sensing region boundary curve two of the pixel-map outside border, curve two lower left to cyan eyeglass, have

Try to achieve I_{RM,Lab_ad}、I_{RM,Lab_bd}Solution；All pixels outside for border, curve two lower left perform this operation；

It is positioned at inside, two dimension sensing region and borderline color dot remains unchanged；After being mapped, right view is at CIEL^*a^*b^*The a of color space^*、b^*Component value I_{RM,Lab_au}、I_{RM,Lab_bu}、I_{RM,Lab_ad}、I_{RM,Lab_bd}, further by a^*、b^*Component value is unified respectively is designated as I_{RM,Lab_a}、I_{RM,Lab_b}；

Step 4, at CIEL^*a^*b^*Left view and right view are carried out brightness regulation by color space respectively；

Red point in note RGB color is Red=(1,0,0), converts it to CIEL^*a^*b^*After color space, calculating the tone value corresponding to this redness point Red is H_Red=41.7515；Take threshold value T=15, to tone value at (H_Red-T, H_Red+ T) color in scope carries out brightness regulation；

To right view, the brightness regulating preceding pixel is I_{R,Lab_L}, the brightness after note adjustment is I_{RM,Lab_L}, have

Wherein, W_rIt is weight coefficient,In formula, W_max=0.4, S_max=50, S_min=40；WithThe respectively tone value of brightness regulation preceding pixel and intensity value；

All pixels of right view are carried out above operation；

To left view, the brightness regulating preceding pixel is I_{L,Lab_L}, the brightness after note adjustment is I_{LM,Lab_L}, have

In formula, W_lFor weight coefficient,Wherein,WithRespectively regulate tone value and the intensity value of preceding pixel；

All pixels of left view are carried out above operation；

Step 5, through above-mentioned coupling map and brightness regulation after, the right view pixel obtained is at CIEL^*a^*b^*The three-component of color space is I_{RM,Lab_L}、I_{RM,Lab_a}、I_{RM,Lab_b}；First, CIE XYZ color space is converted it to:

In formula, I_{RM,XYZ_X}、I_{RM,XYZ_Y}、I_{RM,XYZ_Z}It is that the right view pixel after coupling maps is in the X of CIE XYZ color space, Y, Z component value respectively；f^-1() is the inverse function of f ()；

Pass through A_R ^-1It is shifted to RGB color by CIE XYZ color space contravariant:

All pixels of right view are carried out above operation；

Left view after brightness regulation, the pixel L obtained^*Value is I_{LM,Lab_L}, remember a^*、b^*Component value is I_{LM,Lab_a}、I_{LM,Lab_b}；Using CIE XYZ color space as intermediary, its contravariant is shifted to RGB color:

In formula, I_{LM,XYZ_X}, I_{LM,XYZ_Y}, I_{LM,XYZ_Z}For the X in CIE XYZ color space of the left view after brightness regulation, Y, Z component value, A_LBe according to red eyeglass, the absorption curve function of different-waveband color is calculated, by the transformation matrix of RGB color to CIEXYZ,[X_{L_white}Y_{L_white}Z_{L_white}] it is that CIE XYZ color space corresponds to transformation matrix A_LReference white point,

[X_{L_white}Y_{L_white}Z_{L_white}]^T=A_L×[111]^T

Pass through A_L ^-1It is converted into RGB color by CIE XYZ color space:

If the dark purple anaglyphs pixel of final synthesis is expressed as I at RGB color_A,RGB, tri-channel value respectively I of R, G, B of its correspondence_{A,RGB_R}、I_{A,RGB_G}And I_{A,RGB_B}；By through the right view pixel G of aforementioned processing, B color channel value I_{RM,RGB_G}And I_{RM,RGB_B}Respectively as the G of anaglyphs pixel, B color channel value, by the left view pixel R color channel values I after process_{LM,RGB_R}As the R color channel values of dark purple anaglyphs pixel, [I_{A,RGB_R}I_{A,RGB_G}I_{A,RGB_B}]=[I_{LM,RGB_R}I_{RM,RGB_G}I_{RM,RGB_B}]；Each component is multiplied by the pixel color value I of the 255 dark purple anaglyphses that can calculate final synthesis_A,RGB；

Each pixel of dark purple anaglyphs is performed aforesaid operations, finally gives dark purple anaglyphs.

The invention has the beneficial effects as follows: considered in auto-stereoscopic imaging technique, easily cause visual fatigue visual rivalry phenomenon and cross-color problem, by the left view of stereo image pair and right view are processed respectively, synthesize dark purple anaglyphs.Utilize CIEL^*a^*b^*The perception uniformity of color space, based on tone invariance, at CIEL^*a^*b^*The a of color space^*-b^*In plane, the distribution of color of right view is mated the two-dimentional sensing region mapping to cyan eyeglass, solve owing to part colours is stopped the cross-color problem caused by cyan eyeglass；Change according to tone Yu saturation value, dynamically regulates the brightness of left view and right view, reduces the brightness value of the too high pixel of saturation, reduces the purpose of visual rivalry.It is different from XYZ algorithm and HSV algorithm and exchanges, for cost, the way that visual rivalry phenomenon is improved with huge cross-color for, the present invention considers Problems existing in auto-stereoscopic imaging technique, improving color fidelity and reduce visual rivalry to have reached active balance in improving the stability of stereoscopic vision perception so that the dark purple anaglyphs of synthesis can provide more comfortable stereo vision experience.

Accompanying drawing explanation

Fig. 1 is the method flow diagram of the present invention；

Fig. 2 is matching border, cyan eyeglass two dimension sensing region schematic diagram in CIEL*a*b* color space；

In figure, 1-curve one, 2-curve two.

Detailed description of the invention

Below in conjunction with drawings and Examples, the present invention is further described, and the present invention includes but are not limited to following embodiment.

Step one, left view and the right view of stereo image pair are converted into CIEL by RGB color respectively^*a^*b^*Color space.

First, at RGB color, the left view pixel I to stereo image pair_L,RGBWith right view pixel I_R,RGBTri-channel value of respective R, G, B are normalized so that it is be distributed in [0,1] scope.Again left view pixel and right view pixel are converted into CIEL by RGB color respectively^*a^*b^*Color space, this conversion needs using CIE XYZ color space as intermediary:

In formula, I_{L,RGB_R}、I_{L,RGB_G}、I_{L,RGB_B}Respectively left view pixel I_L,RGBCorresponding tri-channel value of R, G, B, I_{L,XYZ_X}、I_{L,XYZ_Y}、I_{L,XYZ_Z}It is then left view pixel I respectively_L,RGBIn the X of CIE XYZ color space, Y, Z component value.Correspondingly, I_{R,RGB_R}、I_{R,RGB_G}、I_{R,RGB_B}Respectively right view pixel I_R,RGBCorresponding tri-channel value of R, G, B, I_{R,XYZ_X}、I_{R,XYZ_Y}、I_{R,XYZ_Z}It is then that right view pixel is in the X of CIE XYZ color space, Y, Z component value respectively.Cs is the transformation matrix being converted into CIE XYZ color space by RGB color, and value is delivered the accepted value in document:

Recycling CIE XYZ color space and CIEL^*a^*b^*The transformation relation of color space, is converted into CIEL further respectively by the left view pixel and right view pixel that transform to CIE XYZ color space^*a^*b^*Color space:

In formula, I_{L,Lab_L}、I_{L,Lab_a}、I_{L,Lab_b}Respectively left view pixel I_L,RGBAt CIEL^*a^*b^*The L of color space^*、a^*、b^*Component value；I_{R,Lab_L}、I_{R,Lab_a}、I_{R,Lab_b}It is right view pixel I respectively_R,RGBAt CIEL^*a^*b^*The L of color space^*、a^*、b^*Component value.The expression of f () is:

[X_whiteY_whiteZ_white] it is the reference white point of CIE XYZ color space, can be calculated by following formula:

[X_whiteY_whiteZ_white]^T=Cs × [111]^T(7)

It is taken as the unified standard of color reference with reference to white point, ensures the concordance of color measurements and evaluation with this.

Each pixel in left view and right view is carried out above-mentioned computing, thus completing stereo pairs from RGB color to CIEL^*a^*b^*The conversion of color space.

Step 2, calculate cyan eyeglass sensing region.

Color is the human eye visual perception that different wave length visible radiation causes, and is usually the complex light that monochromatic light combines, and color perception is decided by the relative power distribution of different wave length light radiation.Two eyeglasses based on the anaglyphic spectacles of dark purple complementary color are different to the transmitance of different-waveband color.The color filter film that right view uses is cyan, G, B dichromatism synthesize, and when right eye is by eyeglass viewing image, by eyeglass stop, some color can be caused that right eye cannot be observed, thus causing the anaglyphs cross-color finally seen.Here, all cyan eyeglasses that can pass through are observed the color perceived at CIEL by right eye^*a^*b^*The region of Color-spatial distribution is called the three-dimensional perception region of cyan eyeglass, by three-dimensional perception region at a^*-b^*The projection of plane is called two dimension sensing region, and both are referred to as sensing region.In anaglyphs synthesizes, it is necessary to all colours in right view is mapped to the sensing region (including in sensing region and border, sensing region) of cyan eyeglass.Especially, for being positioned at the color outside sensing region, it is necessary to when keeping color constant, to major general, it maps on border, sensing region so that it is color filter film can be passed through and perceived by eyes.

Defining a panchromatic right view comprising all colours, its color is at RGB color three-dimensional matrice C_{256×256×256}Storage, the every one-dimensional element of matrix is the integer from 0 to 255.Calculate this panchromatic right view through after cyan eyeglass at CIEL^*a^*b^*The distributed areas of color space, are cyan eyeglass at CIEL^*a^*b^*The sensing region of color space, its computational methods are as follows:

In order to reduce amount of calculation, utilize the rule of conversion of RGB color and hsv color space, by C_{256×256×256}Replace with the V 100 width HS figure changed from 0 to 99 integer.V is the luminance component in hsv color space, in HS figure to be then V be span during certain constant, using tone H as transverse axis, saturation S is as the 2 d plane picture of the longitudinal axis, tone H axle positive direction level is to the right, span is the integer of 0 to 359, and vertically downward, span is the integer of 0 to 99 to saturation S axle positive direction.If correspondence HS during V=i_iFigure, i is integer and 0≤i≤99, I_{HS_Ri}、I_{HS_Gi}、I_{HS_Bi}Represent HS respectively_iThe color of figure is at tri-channel value of R, G, B of RGB color, I_{RP,XYZ_Xi}、I_{RP,XYZ_Yi}、I_{RP,XYZ_Zi}Represent that it is at CIE XYZ color space tri-component values of corresponding X, Y, Z, then have respectively:

In formula, I_{RP,Lab_Li}、I_{RP,Lab_ai}、I_{RP,Lab_bi}Represent HS respectively_iThe color of figure is at CIEL^*a^*b^*The corresponding L of color space^*、a^*、b^*Component value, A_RBe according to cyan eyeglass, the absorption curve function of different-waveband color is calculated, by the transformation matrix of RGB color to CIEXYZ, its value is delivered the accepted value in document:

[X_{R_white}Y_{G_white}Z_{B_white}] it is that CIE XYZ color space corresponds to transformation matrix A_RReference white point, can be calculated by following formula:

[X_{R_white}Y_{R_white}Z_{R_white}]^T=A_R×[111]^T(11)

According to calculated all I_{RP,Lab_Li}、I_{RP,Lab_ai}、I_{RP,Lab_bi}, at CIEL^*a^*b^*Color space drawing 3 D graphics, obtains the three-dimensional perception region of cyan eyeglass.This sensing region be distributed as a series of point bunch body, be approximately perpendicular to a from spatially observing it^*-b^*Plane.By three-dimensional perception region projection to a^*-b^*Plane, obtains two dimension sensing region, utilizes quadratic polynomial to carry out curve fitting the boundary point bunch of two dimension sensing region, and fitting result is as shown in Figure 2.Top-right edge fitting curve is curve one, and equation is:

b₁=-5.1146 × 10^-4·(a₁)²-0.8724·(a₁)+22.5848(12)

In formula, b₁、a₁The respectively b of curve one^*Component value and a^*Component value.The edge fitting curve of lower left is curve two, and equation is:

b₂=0.0029 (a₂)²-0.89·(a₂)-5.5028(13)

In formula, b₂、a₂The respectively b of curve two^*Component value and a^*Component value.So at a^*-b^*In plane, the region that namely the two-dimentional sensing region of cyan eyeglass is curve one and curve two surrounds.

Step 3, coupling map

When arbitrary right view is converted into CIEL from RGB color^*a^*b^*During color space, major part color dot has transformed to the sensing region of cyan eyeglass, it is no longer necessary to adjust, but some color dot can exceed this region, these colors are perceived in order to enable right eye to observe, can being adjusted on the border, sensing region of cyan eyeglass, this process completes by mating mapping.Correlational study shows, human visual system is processing the final stage of color-aware, the three elements of color are generally used to characterize, the present invention utilizes tone key element therein, color is adjusted so that tone is constant as constraints, so more conform to the human eye perception to color, it is possible to improve the color fidelity of the anaglyphs of synthesis.Select to carry out color adjustment so that tone is constant as constraints, it is meant that coupling maps only need at a^*-b^*Plane carries out, and the color of right view is mapped to CIEL^*a^*b^*The two-dimentional sensing region of cyan eyeglass in color space.

1. coupling maps principle

Coupling mapping needs to follow two principles:

Principle (1): the tone value of all pixels remains unchanged before the mapping afterwards, i.e. tone invariance.In this, as the matching condition of color in mapping process.

Principle (2): a of all pixels of right view^*、b^*Component value should be mapped in the two-dimentional sensing region of cyan eyeglass or on the border of two-dimentional sensing region.

2. carry out coupling using tone invariance as constraints to map

It is transformed into CIEL from RGB color by arbitrary^*a^*b^*The right view distribution of color of color space, the cyan eyeglass two dimension sensing region surrounded with curve one and curve two compares.

The a of border, curve one upper right side exterior pixel after note coupling mapping^*、b^*Component value is I_{RM,Lab_au}、I_{RM,Lab_bu}, tone value is H_RMu, have:

The corresponding a before coupling mapping^*、b^*Component value is I_{R,Lab_au}、I_{R,Lab_bu}, tone value is H_Ru, have:

(1) has H on principle_RMu=H_Ru, it is derived by:

The a of border, curve two lower left exterior pixel after note coupling mapping^*、b^*Component value is I_{RM,Lab_ad}、I_{RM,Lab_bd}, tone value is H_RMd, have:

The corresponding a before coupling mapping^*、b^*Component value is I_{R,Lab_ad}、I_{R,Lab_bd}, tone value is H_Rd, have:

(1) has H on principle_RMd=H_Rd, it is derived by:

For the pixel outside border, curve one upper right side, on principle (2), mapped on the two-dimentional sensing region boundary curve one of cyan eyeglass, i.e. I_{RM,Lab_au}、I_{RM,Lab_bu}Meet formula (12).Had by formula (12) and formula (16):

I can be tried to achieve_{RM,Lab_au}、I_{RM,Lab_bu}Solution.All pixels outside for border, curve one upper right side perform this operation.

For the pixel outside border, curve two lower left, on principle (2), mapped on the two-dimentional sensing region boundary curve two of cyan eyeglass, i.e. I_{RM,Lab_ad}、I_{RM,Lab_bd}Meet formula (13).Had by formula (13) and formula (19):

I can be tried to achieve_{RM,Lab_ad}、I_{RM,Lab_bd}Solution.All pixels outside for border, curve two lower left perform this operation.

It is positioned at inside, two dimension sensing region and borderline color dot remains unchanged.

So far, at a^*-b^*Plane completes the coupling to right view and maps, and after being mapped right view at CIEL^*a^*b^*The a of color space^*、b^*Component value I_{RM,Lab_au}、I_{RM,Lab_bu}、I_{RM,Lab_ad}、I_{RM,Lab_bd}, further by a^*、b^*Component value is unified respectively is designated as I_{RM,Lab_a}、I_{RM,Lab_b}。

Step 4, at CIEL^*a^*b^*Left view and right view are carried out brightness regulation by color space respectively.

In dark purple anaglyphs, if the saturation of red component and brightness are simultaneously too high, visual rivalry phenomenon can be caused fairly obvious, directly influence the level of comfort of human eye viewing anaglyphs.Brightness regulation is carried out to reduce the impact of visual rivalry here by red component color within the specific limits in left view and right view.

Red point in note RGB color is Red=(1,0,0), converts it to CIEL^*a^*b^*After color space, calculating the tone value corresponding to this redness point Red is H_Red=41.7515.According to the summary of experience of many experiments operation, take threshold value T=15, to tone value at (H_Red-T, H_Red+ T) color in scope carries out brightness regulation.

To right view, the brightness regulating preceding pixel is I_{R,Lab_L}, the brightness after note adjustment is I_{RM,Lab_L}, have:

Wherein, W_rBeing weight coefficient, change according to the size of respective pixel color saturation on right view, its expression formula is:

In formula, W_max=0.4, S_max=50, S_min=40, the experiment experience according to delivering HSV algorithm in document is arranged.

H_RWith S_RThe respectively tone value of brightness regulation preceding pixel and intensity value, can be calculated by following formula:

Formula (22) and formula (23) show, when pixel intensity is more high time, and W_rCan be more big accordingly, the brightness I after adjustment_{RM,Lab_L}Can be more little accordingly, avoid the impact of visual rivalry with this；When pixel intensity regulates little, it was shown that itself caused visual rivalry impact is also less.

All pixels of right view are carried out above operation.

Similarly, to left view, the brightness regulating preceding pixel is I_{L,Lab_L}, the brightness after note adjustment is I_{LM,Lab_L}, have:

In formula, W_lFor weight coefficient, changing according to the size of respective pixel color saturation on left view, its effect is similar to W_r, it is embodied as:

Wherein, H_LWith S_LRespectively regulate tone value and the intensity value of preceding pixel, can be calculated by following formula:

All pixels of left view are carried out above operation.

So far, complete the brightness regulation to left view and right view, obtain after brightness regulation left view and right view at CIEL^*a^*b^*The L of color space^*Component value I_{LM,Lab_L}、I_{RM,Lab_L}.By brightness regulation, the color accuracy of red area can be improved on the one hand, it is to avoid excessive cross-color, the brightness disproportionation due to left, view perception can be avoided on the other hand and the visual rivalry that causes.

Step 5, by process after left view and right view by CIEL^*a^*b^*Color space contravariant shifts to RGB color, synthesizes dark purple anaglyphs.

1. right view conversion

After the mapping of above-mentioned coupling and brightness regulation, the right view pixel obtained is at CIEL^*a^*b^*The three-component of color space is I_{RM,Lab_L}、I_{RM,Lab_a}、I_{RM,Lab_b}.First, CIE XYZ color space is converted it to:

In formula, I_{RM,XYZ_X}、I_{RM,XYZ_Y}、I_{RM,XYZ_Z}It is that the right view pixel after coupling maps is in the X of CIE XYZ color space, Y, Z component value respectively.f^-1() is the inverse function of f (), and its expression is:

Again through A_R ^-1It is shifted to RGB color by CIE XYZ color space contravariant:

All pixels of right view are carried out above operation.

2. left view transformation

Left view after brightness regulation, the pixel L obtained^*Value is I_{LM,Lab_L}, remember a^*、b^*Component value is I_{LM,Lab_a}、I_{LM,Lab_b}.Do not carry out coupling to map because of left view, a^*、b^*Component is not changed in, therefore I_{LM,Lab_a}=I_{R,Lab_a}, I_{LM,Lab_b}=I_{R,Lab_b}.Similarly, using CIE XYZ color space as intermediary, its contravariant is shifted to RGB color:

In formula, I_{LM,XYZ_X}, I_{LM,XYZ_Y}, I_{LM,XYZ_Z}For the X in CIE XYZ color space of the left view after brightness regulation, Y, Z component value.A_LBe according to red eyeglass, the absorption curve function of different-waveband color is calculated, by the transformation matrix of RGB color to CIEXYZ, its value is delivered the accepted value in document:

[X_{L_white}Y_{L_white}Z_{L_white}] it is that CIE XYZ color space corresponds to transformation matrix A_LReference white point, can be calculated by following formula:

[X_{L_white}Y_{L_white}Z_{L_white}]^T=A_L×[111]^T(35)

Pass through A_L ^-1It is converted into RGB color by CIE XYZ color space:

3. the synthesis of dark purple anaglyphs

If the dark purple anaglyphs pixel of final synthesis is expressed as I at RGB color_A,RGB, tri-channel value respectively I of R, G, B of its correspondence_{A,RGB_R}、I_{A,RGB_G}And I_{A,RGB_B}.By through the right view pixel G of aforementioned processing, B color channel value I_{RM,RGB_G}And I_{RM,RGB_B}Respectively as the G of anaglyphs pixel, B color channel value, by the left view pixel R color channel values I after process_{LM,RGB_R}R color channel values as dark purple anaglyphs pixel.Then have:

[I_{A,RGB_R}I_{A,RGB_G}I_{A,RGB_B}]=[I_{LM,RGB_R}I_{RM,RGB_G}I_{RM,RGB_B}](37)

Each component is multiplied by the pixel color value I of the 255 dark purple anaglyphses that can calculate final synthesis_A,RGB.Each pixel of dark purple anaglyphs is performed aforesaid operations, finally gives dark purple anaglyphs.

So far, by above five steps, the synthesis of dark purple anaglyphs is completed.

Following four embodiment is AMDAthlon (tm) IIX2245Processor at CPU, dominant frequency is 3.01GHz, inside save as and the computer of 2GB carries out, film source is color solid image pair, input is left view and the right view of color solid image pair, corresponding programming language is matlab, it is achieved software platform be matlab2010a.

Embodiment finds on image four characteristic points that geometric properties is obvious, be prone to location, the arbitrary view of anaglyphs and stereo image pair (selecting right view here) is determined 4 stack features points pair with four Feature point correspondence respectively, detect the RGB color value of 4 stack features points pair, utilize the color fidelity performance of " aberration percentage ratio " index verification algorithm.At RGB color, the aberration percentage ratio defining every stack features point pair is:

Wherein, R₁、G₁、B₁Respectively tri-component values of R, G, B of characteristic point on anaglyphs；R₀、G₀、B₀Respectively tri-component values of R, G, B of character pair point on the right view of stereo pairs.Aberration percentage ratio is more big, it was shown that the color fidelity of the anaglyphs of algorithm synthesis is more poor；Aberration percentage ratio is more little, then corresponding color fidelity is more good.

It is respectively adopted the algorithm synthesis dark purple anaglyphs that XYZ algorithm, HSV algorithm and the present invention propose, is calculated analyzing contrast.Due to empirical composition algorithm and the visual rivalry phenomenon of the anaglyphs that Lab algorithm synthesizes compared with above-mentioned three kinds of algorithms significantly greater, do not improve at the visual rivalry that the present invention is targeted within the scope of the discussion of algorithm, therefore be not involved in contrast.

Embodiment 1:

Input size is left view and the right view of " duck " color solid image pair of 400 × 300 pixels, is respectively adopted XYZ algorithm, HSV algorithm and the present invention and synthesizes dark purple anaglyphs.Obvious geometric properties according to right view non-occluded area, choose four characteristic points, be designated as a little 1, point 2, point 3 and point 4, determine 4 stack features point pair on right view and dark purple anaglyphs according to this, record R, G, B three-component value of 4 stack features points pair, calculating aberration percentage ratio, result is as shown in table 1:

The color contrast of each dark purple anaglyphs composition algorithm of table 1 " duck " image

As it can be seen from table 1 the anaglyphs of present invention synthesis is less by 9.03% than XYZ algorithm at the aberration percentage ratio of point 1 calculating, less by 11.18% than HSV algorithm；Aberration percentage difference in point 2 calculating reaches minimum, and its value is less by 10.81% than XYZ algorithm, less by 12.32% than HSV algorithm；Less by 4.54% than XYZ algorithm at the aberration percentage ratio of point 3 calculating, less by 1.42% than HSV algorithm；Less by 3.08% than XYZ algorithm at the aberration percentage ratio of point 4 calculating, less by 2.79% than HSV algorithm.Data above shows, the dark purple anaglyphs of present invention synthesis can improve the fidelity simultaneously effective improving color of visual rivalry.

Embodiment 2:

Input size is left view and the right view of " memorial archway " color solid image pair of 800 × 600 pixels, is respectively adopted XYZ algorithm, HSV algorithm and the present invention and synthesizes dark purple anaglyphs.Obvious geometric properties according to right view non-occluded area, choose four characteristic points, be designated as a little 1, point 2, point 3 and point 4, determine 4 stack features point pair on right view and dark purple anaglyphs according to this, record R, G, B three-component value of 4 stack features points pair, calculating aberration percentage ratio, result is as shown in table 2:

The color contrast of each dark purple anaglyphs composition algorithm of table 2 " memorial archway " image

From table 2 it can be seen that the anaglyphs of present invention synthesis reaches minimum in the aberration percentage difference of point 1 calculating, its value is less by 38.02% than XYZ algorithm, less by 26.15% than HSV algorithm；Less by 2.2% than XYZ algorithm at the aberration percentage ratio of point 2 calculating, less by 1.29% than HSV algorithm；Less by 14.13% than XYZ algorithm at the aberration percentage ratio of point 3 calculating, less by 17.7% than HSV algorithm；Less by 14.83% than XYZ algorithm at the aberration percentage ratio of point 4 calculating, less by 1.9% than HSV algorithm.Data above shows, the region that the dark purple anaglyphs color fidelity that synthesizes at XYZ algorithm and HSV algorithm is relatively low, the dark purple anaglyphs of present invention synthesis is while improving visual rivalry, and the effect improving color fidelity is especially apparent.

Embodiment 3:

Input size is left view and the right view of " superman " color solid image pair of 420 × 330 pixels, is respectively adopted XYZ algorithm, HSV algorithm and the present invention and synthesizes dark purple anaglyphs.Obvious geometric properties according to right view non-occluded area, choose four characteristic points, be designated as a little 1, point 2, point 3 and point 4, determine 4 stack features point pair on right view and dark purple anaglyphs according to this, record R, G, B three-component value of 4 stack features points pair, calculating aberration percentage ratio, result is as shown in table 3:

The color contrast of each dark purple anaglyphs composition algorithm of table 3 " superman " image

From table 3 it can be seen that the anaglyphs of present invention synthesis reaches minimum in the aberration percentage difference of point 1 calculating, its value is less by 27.26% than XYZ algorithm, less by 31.04% than HSV algorithm；Less by 19.04% than XYZ algorithm at the aberration percentage ratio of point 2 calculating, less by 22.85% than HSV algorithm；Less by 8.04% than XYZ algorithm at the aberration percentage ratio of point 3 calculating, less by 13.43% than HSV algorithm；Less by 2.75% than XYZ algorithm at the aberration percentage ratio of point 4 calculating, less by 4.4% than HSV algorithm.Data above shows, the dark purple anaglyphs of present invention synthesis can improve the fidelity simultaneously effective improving color of visual rivalry.

Embodiment 4:

Input size is left view and the right view of " ocean " color solid image pair of 480 × 220 pixels, is respectively adopted XYZ algorithm, HSV algorithm and the present invention and synthesizes dark purple anaglyphs.Obvious geometric properties according to right view non-occluded area, choose four characteristic points, be designated as a little 1, point 2, point 3 and point 4, determine 4 stack features point pair on right view and dark purple anaglyphs according to this, record R, G, B three-component value of 4 stack features points pair, calculating aberration percentage ratio, result is as shown in table 4:

The color contrast of each dark purple anaglyphs composition algorithm of table 4 " ocean " image

From table 4, it can be seen that the anaglyphs of present invention synthesis reaches minimum in the aberration percentage difference of point 1 calculating, its value is less by 14.68% than XYZ algorithm, less by 17.39% than HSV algorithm；Less by 7.76% than XYZ algorithm at the aberration percentage ratio of point 2 calculating, less by 22.07% than HSV algorithm；Less by 5.47% than XYZ algorithm at the aberration percentage ratio of point 3 calculating, less by 9.85% than HSV algorithm；Less by 5.41% than XYZ algorithm at the aberration percentage ratio of point 4 calculating, less by 8.44% than HSV algorithm.Data above shows, the dark purple anaglyphs of present invention synthesis can improve the fidelity simultaneously effective improving color of visual rivalry.

Above example describes simultaneously, and in the process of synthesis anaglyphs, the distortion level of different colors is different.Generally, Lycoperdon polymorphum Vitt and green distortion are less, and red and blue distortion is bigger.

Claims (1)

1. the anaglyphs visual rivalry ameliorative way improving color fidelity, it is characterised in that comprise the steps:

Step one, at RGB color, the left view pixel I to stereo image pair_L,RGBWith right view pixel I_R,RGBTri-channel value of respective R, G, B are normalized so that it is be distributed in [0,1] scope；As intermediary, left view pixel and right view pixel are converted into CIEL by RGB color respectively using CIE XYZ color space^*a^*b^*Color space, $\left[\begin{array}{c}{I}_{L,XYZ\_X}\\ I_{L,XYZ\_Y}\\ I_{L,XYZ\_Z}\end{array}\right]=Cs\×\left[\begin{array}{c}{I}_{L,RGB\_R}\\ I_{L,RGB\_G}\\ I_{L,RGB\_B}\end{array}\right],\left[\begin{array}{c}{I}_{R,XYZ\_X}\\ I_{R,XYZ\_Y}\\ I_{R,XYZ\_Z}\end{array}\right]=Cs\×\left[\begin{array}{c}{I}_{R,RGB\_R}\\ I_{R,RGB\_G}\\ I_{R,RGB\_B}\end{array}\right],$ In formula, I_{L,RGB_R}、I_{L,RGB_G}、I_{L,RGB_B}Respectively left view pixel I_L,RGBCorresponding tri-channel value of R, G, B, I_{L,XYZ_X}、I_{L,XYZ_Y}、I_{L,XYZ_Z}It is then left view pixel I respectively_L,RGBAt the X of CIE XYZ color space, Y, Z component value, I_{R,RGB_R}、I_{R,RGB_G}、I_{R,RGB_B}Respectively right view pixel I_R,RGBCorresponding tri-channel value of R, G, B, I_{R,XYZ_X}、I_{R,XYZ_Y}、I_{R,XYZ_Z}It is then that right view pixel is in the X of CIE XYZ color space, Y, Z component value respectively；Cs is the transformation matrix being converted into CIE XYZ color space by RGB color, $Cs=\left[\begin{array}{ccc}0.4243& 0.3105& 0.1657\\ 0.2492& 0.6419& 0.1089\\ 0.0265& 0.1225& 0.8614\end{array}\right];$

Utilize CIE XYZ color space and CIEL^*a^*b^*The transformation relation of color space, is converted into CIEL further respectively by the left view pixel and right view pixel that transform to CIE XYZ color space^*a^*b^*Color space, $\left\{\begin{array}{c}{I}_{L,Lab\_L}=116\·f\left(\frac{I_{L,XYZ\_Y}}{Y_{white}}\right)-16\\ I_{L,Lab\_a}=500\·\[f\left(\frac{I_{L,XYZ\_X}}{X_{white}}\right)-f\left(\frac{I_{L,XYZ\_Y}}{Y_{white}}\right)\]\\ I_{L,Lab\_b}=200\·\[f\left(\frac{I_{L,XYZ\_Y}}{Y_{white}}\right)-f\left(\frac{I_{L,XYZ\_Z}}{Z_{white}}\right)\]\end{array}\right.,$ $\left\{\begin{array}{c}{I}_{R,Lab\_L}=116\·f\left(\frac{I_{R,XYZ\_Y}}{Y_{white}}\right)-16\\ I_{R,Lab\_a}=500\·\[f\left(\frac{I_{R,XYZ\_X}}{X_{white}}\right)-f\left(\frac{I_{R,XYZ\_Y}}{Y_{white}}\right)\]\\ I_{R,Lab\_b}=200\·\[f\left(\frac{I_{R,XYZ\_Y}}{Y_{white}}\right)-f\left(\frac{I_{R,XYZ\_Z}}{Z_{white}}\right)\]\end{array}\right.,$ In formula, I_{L,Lab_L}、I_{L,Lab_a}、I_{L,Lab_b}Respectively left view pixel I_L,RGBAt CIEL^*a^*b^*The L of color space^*、a^*、b^*Component value；I_{R,Lab_L}、I_{R,Lab_a}、I_{R,Lab_b}It is right view pixel I respectively_R,RGBAt CIEL^*a^*b^*The L of color space^*、a^*、b^*Component value, the expression formula of f () is[X_whiteY_whiteZ_white] it is the reference white point of CIE XYZ color space, [X_whiteY_whiteZ_white]^T=Cs × [111]^T；

Each pixel in left view and right view is carried out above-mentioned computing, thus completing stereo pairs from RGB color to CIEL^*a^*b^*The conversion of color space；

Step 2, one panchromatic right view comprising all colours of definition, its color is at RGB color three-dimensional matrice C_{256×256×256}Storage, the every one-dimensional element of matrix is the integer from 0 to 255；Calculate this panchromatic right view through after cyan eyeglass at CIEL^*a^*b^*The distributed areas of color space, are cyan eyeglass at CIEL^*a^*b^*The sensing region of color space, its computational methods are as follows:

By C_{256×256×256}Replace with the V 100 width HS figure changed from 0 to 99 integer, V is the luminance component in hsv color space, HS figure is V when being certain constant in span, using tone H as transverse axis, saturation S is as the 2 d plane picture of the longitudinal axis, and to the right, span is the integer of 0 to 359 to tone H axle positive direction level, vertically downward, span is the integer of 0 to 99 to saturation S axle positive direction；If correspondence HS during V=i_iFigure, i is integer and 0≤i≤99, I_{HS_Ri}、I_{HS_Gi}、I_{HS_Bi}Represent HS respectively_iThe color of figure is at tri-channel value of R, G, B of RGB color, I_{RP,XYZ_Xi}、I_{RP,XYZ_Yi}、I_{RP,XYZ_Zi}Represent that it is at CIE XYZ color space tri-component values of corresponding X, Y, Z respectively, then $\left[\begin{array}{c}{I}_{RP,XYZ\_Xi}\\ I_{RP,XYZ\_Yi}\\ I_{RP,XYZ\_Zi}\end{array}\right]=A_R\×\left[\begin{array}{c}{I}_{HS\_Ri}\\ I_{HS\_Gi}\\ I_{HS\_Bi}\end{array}\right],\left\{\begin{array}{c}{L}_{RP,Lab\_Li}=116\·f\left(\frac{I_{RP,XYZ\_Yi}}{Y_{R\_white}}\right)-16\\ I_{RP,Lab\_ai}=500\·\[f\left(\frac{I_{RP,XYZ\_Xi}}{X_{R\_white}}\right)-f\left(\frac{I_{RP,XYZ\_Yi}}{Y_{R\_white}}\right)\]\\ I_{RP,Lab\_bi}=200\·\[f\left(\frac{I_{RP,XYZ\_Yi}}{Y_{R\_white}}\right)-f\left(\frac{I_{RP,XYZ\_Zi}}{Z_{R\_white}}\right)\]\end{array}\right.,$ In formula, I_{RP,Lab_Li}、I_{RP,Lab_ai}、I_{RP,Lab_bi}Represent HS respectively_iThe color of figure is at CIEL^*a^*b^*The corresponding L of color space^*、a^*、b^*Component value, $A_R=\left[\begin{array}{ccc}0.0153& 0.1092& 0.1171\\ 0.0176& 0.3088& 0.0777\\ 0.0201& 0.1016& 0.6546\end{array}\right]$ Be according to cyan eyeglass, the absorption curve function of different-waveband color is calculated, by the transformation matrix of RGB color to CIEXYZ, [X_{R_white}Y_{G_white}Z_{B_white}] it is that CIE XYZ color space corresponds to transformation matrix A_RReference white point, [X_{R_white}Y_{R_white}Z_{R_white}]^T=A_R×[111]^T；

According to calculated all I_{RP,Lab_Li}、I_{RP,Lab_ai}、I_{RP,Lab_bi}, at CIEL^*a^*b^*Color space drawing 3 D graphics, obtains the three-dimensional perception region of cyan eyeglass.This sensing region be distributed as a series of point bunch body, be approximately perpendicular to a from spatially observing it^*-b^*Plane.By three-dimensional perception region projection to a^*-b^*Plane, obtains the region that namely two dimension sensing region is curve one and curve two surrounds, and the equation of curve one is b₁=-5.1146 × 10^-4·(a₁)²-0.8724·(a₁)+22.5848, the equation of curve two is b₂=0.0029 (a₂)²-0.89·(a₂)-5.5028, in formula, b₁、a₁The respectively b of curve one^*Component value and a^*Component value, b₂、a₂The respectively b of curve two^*Component value and a^*Component value；

Step 3, at a^*-b^*Plane carries out coupling and maps, and the color of right view is mapped to CIEL^*a^*b^*The two-dimentional sensing region of cyan eyeglass in color space；

The a of border, curve one upper right side exterior pixel after note coupling mapping^*、b^*Component value is I_{RM,Lab_au}、I_{RM,Lab_bu}, tone value isA before Corresponding matching mapping^*、b^*Component value I_{R,Lab_au}、I_{R,Lab_bu}, tone value $H_{Ru}=\frac{180}{\mathrm{\π}}\·arctan\left(\frac{I_{R,Lab\_bu}}{I_{R,Lab\_au}}\right);$ It is derived by $\frac{I_{RM,Lab\_bu}}{I_{RM,Lab\_au}}=\frac{I_{R,Lab\_bu}}{I_{R,Lab\_au}};$

The a of border, curve two lower left exterior pixel after note coupling mapping^*、b^*Component value is I_{RM,Lab_ad}、I_{RM,Lab_bd}, tone valueA before Corresponding matching mapping^*、b^*Component value I_{R,Lab_ad}、I_{R,Lab_bd}, tone value $H_{Rd}=\frac{180}{\mathrm{\π}}\·arctan\left(\frac{I_{R,Lab\_bd}}{I_{R,Lab\_ad}}\right);$ It is derived by $\frac{I_{RM,Lab\_bd}}{I_{RM,Lab\_ad}}=\frac{I_{R,Lab\_bd}}{I_{R,Lab\_ad}};$

By on the two-dimentional sensing region boundary curve one of the pixel-map outside border, curve one upper right side to cyan eyeglass, have

$\left\{\begin{array}{c}\frac{I_{RM,Lab\_bu}}{I_{RM,Lab\_au}}=\frac{I_{R,Lab\_bu}}{I_{R,Lab\_au}}\\ I_{RM,Lab\_bu}=-5.1146\×{10}^{-4}\·{\left(I_{RM,Lab\_au}\right)}^2-0.8724\·I_{RM,Lab\_au}+22.5848\end{array}\right.$

Try to achieve I_{RM,Lab_au}、I_{RM,Lab_bu}Solution；All pixels outside for border, curve one upper right side perform this operation；

By on the two-dimentional sensing region boundary curve two of the pixel-map outside border, curve two lower left to cyan eyeglass, have

$\left\{\begin{array}{c}\frac{I_{RM,Lab\_bd}}{I_{RM,Lab\_ad}}=\frac{I_{R,Lab\_bd}}{I_{R,Lab\_ad}}\\ I_{RM,Lab\_bd}=0.0029\·{\left(I_{RM,Lab\_ad}\right)}^2-0.89\·I_{RM,Lab\_ad}-5.50288\end{array}\right.$

Try to achieve I_{RM,Lab_ad}、I_{RM,Lab_bd}Solution；All pixels outside for border, curve two lower left perform this operation；

It is positioned at inside, two dimension sensing region and borderline color dot remains unchanged；After being mapped, right view is at CIEL^*a^*b^*The a of color space^*、b^*Component value I_{RM,Lab_au}、I_{RM,Lab_bu}、I_{RM,Lab_ad}、I_{RM,Lab_bd}, further by a^*、b^*Component value is unified respectively is designated as I_{RM,Lab_a}、I_{RM,Lab_b}；

Step 4, at CIEL^*a^*b^*Left view and right view are carried out brightness regulation by color space respectively；

Red point in note RGB color is Red=(1,0,0), converts it to CIEL^*a^*b^*After color space, calculating the tone value corresponding to this redness point Red is H_Red=41.7515；Take threshold value T=15, to tone value at (H_Red-T, H_Red+ T) color in scope carries out brightness regulation；

To right view, the brightness regulating preceding pixel is I_{R,Lab_L}, the brightness after note adjustment is I_{RM,Lab_L}, have

Wherein, W_rIt is weight coefficient,In formula, W_max=0.4, S_max=50, S_min=40； $H_R=\frac{180}{\mathrm{\π}}\·arctan\left(\frac{I_{R,Lab\_b}}{I_{R,Lab\_a}}\right)$ With $S_R=\sqrt{{\left(I_{R,Lab\_a}\right)}^2+{\left(I_{R,Lab\_b}\right)}^2}$ The respectively tone value of brightness regulation preceding pixel and intensity value；

All pixels of right view are carried out above operation；

To left view, the brightness regulating preceding pixel is I_{L,Lab_L}, the brightness after note adjustment is I_{LM,Lab_L}, have

In formula, W_lFor weight coefficient,Wherein, $H_L=\frac{180}{\mathrm{\π}}\·arctan\left(\frac{I_{L,Lab\_b}}{I_{L,Lab\_a}}\right)$ With $S_L=\sqrt{{\left(I_{L,Lab\_a}\right)}^2+{\left(I_{L,Lab\_b}\right)}^2}$ Respectively regulate tone value and the intensity value of preceding pixel；All pixels of left view are carried out above operation；

Step 5, through above-mentioned coupling map and brightness regulation after, the right view pixel obtained is at CIEL^*a^*b^*The three-component of color space is I_{RM,Lab_L}、I_{RM,Lab_a}、I_{RM,Lab_b}；First, CIE XYZ color space is converted it to:

$\left\{\begin{array}{c}{I}_{RM,XYZ\_X}=X_{R\_white}\·f^{-1}(\frac{I_{RM,Lab\_L}+16}{116}+\frac{I_{RM,Lab\_a}}{500})\\ I_{RM,XYZ\_Y}=Y_{R\_white}\·f^{-1}\left(\frac{I_{RM,Lab\_L}+16}{116}\right)\\ I_{RM,XYZ\_Z}=Z_{R\_white}\·f^{-1}(\frac{I_{RM,Lab\_L}+16}{116}-\frac{I_{RM,Lab\_b}}{200})\end{array}\right.$

In formula, I_{RM,XYZ_X}、I_{RM,XYZ_Y}、I_{RM,XYZ_Z}It is that the right view pixel after coupling maps is in the X of CIE XYZ color space, Y, Z component value respectively；f^-1() is the inverse function of f ()；

Pass through A_R ^-1It is shifted to RGB color by CIE XYZ color space contravariant:

$\left[\begin{array}{c}{I}_{RM,RGB\_R}\\ I_{RM,RGB\_G}\\ I_{RM,RGB\_B}\end{array}\right]={A_B}^{-1}\×\left[\begin{array}{c}{I}_{RM,XYZ\_X}\\ I_{RM,XYZ\_Y}\\ I_{RM,XYZ\_Z}\end{array}\right];$

All pixels of right view are carried out above operation；

Left view after brightness regulation, the pixel L obtained^*Value is I_{LM,Lab_L}, remember a^*、b^*Component value is I_{LM,Lab_a}、I_{LM,Lab_b}；Using CIE XYZ color space as intermediary, its contravariant is shifted to RGB color:

$\left\{\begin{array}{c}{I}_{LM,XYZ\_X}=X_{L\_white}\·f^{-1}(\frac{I_{LM,Lab\_L}+16}{116}+\frac{I_{Lm,Lab\_a}}{500}))\\ I_{LM,XYZ\_Y}=Y_{L\_white}\·f^{-1}\left(\frac{I_{LM,Lab\_L}+16}{116}\right)\\ I_{LM,XYZ\_Z}=Z_{L\_white}\·f^{-1}(\frac{I_{LM,Lab\_L}+16}{116}-\frac{I_{LM,Lab\_b}}{200})\end{array}\right.$

In formula, I_{LM,XYZ_X}, I_{LM,XYZ_Y}, I_{LM,XYZ_Z}For the X in CIE XYZ color space of the left view after brightness regulation, Y, Z component value, A_LBe according to red eyeglass, the absorption curve function of different-waveband color is calculated, by the transformation matrix of RGB color to CIEXYZ, $A_L=\left[\begin{array}{ccc}0.1840& 0.0179& 0.0048\\ 0.0876& 0.0118& 0.0018\\ 0.0005& 0.0012& 0.0159\end{array}\right],$ [X_{L_white}Y_{L_white}Z_{L_white}] it is that CIE XYZ color space corresponds to transformation matrix A_LReference white point,

[X_{L_white}Y_{L_white}Z_{L_white}]^T=A_L×[111]^T

Pass through A_L ^-1It is converted into RGB color by CIE XYZ color space:

$\left[\begin{array}{c}{I}_{LM,RGB\_R}\\ I_{LM,RGB\_G}\\ I_{LM,RGB\_B}\end{array}\right]={A_L}^{-1}\×\left[\begin{array}{c}{I}_{LM,XYZ\_X}\\ I_{LM,XYZ\_Y}\\ I_{LM,XYZ\_Z}\end{array}\right];$

If the dark purple anaglyphs pixel of final synthesis is expressed as I at RGB color_A,RGB, tri-channel value respectively I of R, G, B of its correspondence_{A,RGB_R}、I_{A,RGB_G}And I_{A,RGB_B}；By through the right view pixel G of aforementioned processing, B color channel value I_{RM,RGB_G}And I_{RM,RGB_B}Respectively as the G of anaglyphs pixel, B color channel value, by the left view pixel R color channel values I after process_{LM,RGB_R}As the R color channel values of dark purple anaglyphs pixel, [I_{A,RGB_R}I_{A,RGB_G}I_{A,RGB_B}]=[I_{LM,RGB_R}I_{RM,RGB_G}I_{RM,RGB_B}]；Each component is multiplied by the pixel color value I of the 255 dark purple anaglyphses that can calculate final synthesis_A,RGB；

Each pixel of dark purple anaglyphs is performed aforesaid operations, finally gives dark purple anaglyphs.