METHOD FOR COLORING AN INFRARED IMAGE

The CIE CAM02-UCS space with a parametric trajectory ensures linear perception in colorizing infrared images, addressing individual perception variations to enhance detail visibility and temperature differentiation.

FR3142024B1Active Publication Date: 2024-11-08LYNRED
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
FR2022011811
Authority / Receiving Office
FR · FR
Patent Type
Patents
Current Assignee / Owner
Filing Date
2022-11-14
Publication Date
2024-11-08
Estimated Expiration
2042-11-14

AI Technical Summary

Technical Problem

Existing methods for colorizing infrared images fail to account for individual perception, resulting in non-linear color palettes that may not effectively highlight details for specific users, complicating the interpretation of temperature differences.

Method used

A color palette defined in the CIE CAM02-UCS space using a parametric trajectory with a polar angle vector, ensuring linear perception by projecting hue and chrominance as ellipses and luminance as straight lines, allowing users to adjust parameters for personal preferences.

Benefits of technology

Enables personalized and effective colorization of infrared images, enhancing the visibility of temperature differences and object salience by aligning with individual perception criteria.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure 00000019_0000
    Figure 00000019_0000
  • Figure 00000019_0001
    Figure 00000019_0001
  • Figure 00000019_0002
    Figure 00000019_0002
Patent Text Reader

Abstract

This method for colouring an infrared image (11) comprises the following steps: – acquisition of an infrared image (11) obtained from an infrared sensor (10) comprising an array of elementary detectors, said infrared image (11) comprising, for each elementary detector, a pixel whose value is coded in grey levels (NG) corresponding to a value representing an infrared radiation received by said elementary detector; and – generation of a coloured image (13) by associating to each pixel value coded in grey levels (NG) a pixel value coded in a colour space by means of a colour palette (12); said colour palette (12) being defined in the CIE CAM02-UCS space by a palette generation module (20) using a parametric trajectory in which a parameter is a polar angle vector (Φ) of the parametric trajectory, the latter being defined such that:.The parametric trajectory projected onto a hue and chrominance plane (O, a*, b*) follows one or two ellipses; and a dynamic range of perceived luminance (j*) with respect to gray level values ​​(NG) follows one or two straight lines. Figure for the abstract: Fig 1.
Need to check novelty before this filing date? Find Prior Art

Description

Title of the invention: METHOD FOR COLORING AN INFRARED IMAGE Technical field

[0001] The present invention relates to the field of the detection of electromagnetic radiation and, more specifically, to the detection of infrared radiation.

[0002] The invention relates more specifically to the problem of displaying an infrared image, that is to say an image resulting from the detection of infrared radiation. More particularly, it aims to colorize an infrared image to improve the interpretation of the infrared image by the human brain.

[0003] Thus, the invention can be implemented in many fields for which infrared images are currently used, such as the fields of aerospace, security, defense, transportation, thermography, industrial inspection, building inspection, leisure, health, etc. State of the art

[0004] In the field of health, infrared detectors have seen great development since the COVID 19 pandemic. Indeed, people suffering from this disease typically have a rise in temperature and infrared detectors make it possible to obtain an image illustrating the temperature of objects present in a scene. Thus, infrared detectors are used in particular to identify the temperature of people entering an enclosure in order to detect whether a person has a high temperature, and therefore a risk of being positive for COVID 19.

[0005] Infrared detectors conventionally make it possible to obtain an image restored in gray levels, in which each pixel represents the quantity of radiation captured by the elementary detectors forming the pixels of the infrared image.

[0006] However, this grayscale representation makes it difficult to visually identify the temperature difference between two objects in a scene with similar temperatures.

[0007] For example, in the case of detecting people carrying COVID 19, this grayscale representation is often insufficient for an operator to quickly detect a person with a high temperature in a group of people.

[0008] It is therefore sought to improve this representation by coloring the representation of the quantity of radiation captured by the elementary detectors forming the pixels of the infrared image.

[0009] This research is motivated in particular by the fact that the human eye can only perceive a limited number of shades of gray, between 120 and 180 shades for a trained Western person according to studies, while a person can perceive up to 2 million colors. Thus, by choosing the colors carefully, it is possible to significantly improve the perception of the details of an infrared image.

[0010] For this purpose, it is known from documents WO 2016 / 179050, US 10,298,859 or WO 2014 / 200586 to transform an infrared image in grayscale into a colorized image by means of the application of a color palette. A color palette makes it possible to associate, for each pixel value coded in grayscale, a pixel value coded in a color space.

[0011] To define a color space, we consider the scene light spectrum W(X) in units of power, with X the wavelength in micrometers. In the case of a real scene in visible light, W(X) is the reflected, scattered or transmitted fraction of the scene illuminated by an external source of given spectral length, preferably a white source, i.e. a wide band in the visible. In the case of a display by a screen, W(X) is an additive linear combination of the emission spectra of a triplet of elementary sources, red, green and blue pixels.

[0012] The human eye perceives brightness and color respectively through the rods and cones present in the retina. In low light conditions, typically in scotopic vision, the rods provide a single spectral response. In good light conditions, in photopic vision, the cones participate in the perception of color, and are classified into three families: short, medium and long, which present distinct spectral responses centered on short, medium and long wavelengths.

[0013] Thus, for the human eye, color is characterized by a simple scalar triplet, and not a complete scene spectrum. This argument helps explain how a color can be faithfully reproduced via a screen made up of three elementary sources of red, green, and blue spectral lengths.

[0014] In the sRGB space for “standard Reg-Green-Blue” in Anglo-Saxon literature, a color is represented by the triplet of red, green and blue colors, generally coded on a depth of three times 8 bits, or 24 bits.

[0015] To obtain a more accurate representation of colors, other spaces have been defined over time. For example, the CIE XYZ space is a coordinate system for color space that the International Commission on Illumination (CIE) defined in 1931. It is derived from the sRGB space and postulates that the relationships between physical quantities and perception are linear. This approximation places all the points representing colors in a three-dimensional vector space. To define this CIE XYZ space, colori- metrics were required by asking people to adjust the proportions of three experimental primary colors so as to obtain a color patch identical to the one being evaluated. The sum of the "X", "Y" and "Z" components represents the absolute luminance, but not the actually perceived luminance.

[0016] The International Commission on Illumination (CIE) also defined the CIE LAB space in 1976, based on the CIE XYZ space. This is a colorimetric space particularly used for the characterization of surface colors.

[0017] Three quantities characterize the colors: the lightness L, which derives from the luminance of the surface, and two parameters a and b which express the deviation of the color from that of a gray surface of the same lightness. The existence of a gray, uncolored, achromatic surface implies explicitly indicating the composition of the light which illuminates the colored surface. This "illuminant" is often standardized daylight.

[0018] The major defect of the CIE LAB space is that it does not correctly model the human perception of luminance and hue. That is to say, any displacement of an elementary Euclidean length in the CIE LAB space does not necessarily lead to the same color perception distance, particularly in terms of luminance and hue. This is explained by the fact that this model does not take into account the principles of perceptual adaptation of chromaticity, such as the local adaptation of luminance and hue.

[0019] More recently, in 2002, the International Commission on Illumination (CIE) defined the CIE CAM02 space to calculate the mathematical correlations existing between the six technical dimensions linked to the appearance of color: brightness, luminance, colorization, chrominance, saturation and hue.

[0020] This model is defined from the quantities X, Y and Z of the CIE XYZ space taking into account the visualization of a reference stimulus. This stimulus corresponds to a reference white point. The CIE CAM02 model also takes into consideration the background on which the stimulus is observed, information on the luminance conditions of the stimulus's surroundings, as well as information on whether or not the luminance conditions are taken into account by the observer of the stimulus. The model makes it possible to predict the appearance of the stimulus color with these attributes or to calculate the corresponding color under distinct observation conditions.

[0021] There are also other color spaces, for example the HSV space, for “Hue-Saturation-Value” in Anglo-Saxon literature, or the HSL space, for “Hue-Saturation-Luminance”.

[0022] For example, WO 2014 / 200586 proposes a method for generating color palettes based on interpolation in HSL space. From a series of predefined hue values, the palette is constructed by constant luminance, fixed saturation and linear interpolation on the hue. Depending on the type of scene, this This method allows the use of suitable color ranges, for example a red-orange-yellow palette for desert scenes, or a yellow-green-blue palette for forest scenes.

[0023] In document WO 2016 / 179050, color palettes are generated by interpolation in the sRGB space from a series of basic colors, corresponding to interpolation nodes.

[0024] Document US 10,298,859 presents colorization solutions in the sRGB, CIE XYZ or CIE LAB space, mainly using predefined color palettes for specific applications. In one embodiment, a color palette is generated from the minimum and maximum gray level values ​​of an infrared image. This color palette is therefore adapted to the dynamics of the observed scene.

[0025] Thus, a scene with low thermal contrast is represented with low hue dynamics, which is compensated by adapted luminance dynamics. This document also proposes to increase the perceived contrast by using a luminance filter which exhibits oscillations.

[0026] Whatever the method used, the objective is to obtain a colorization palette on a monochrome image, by following a sequence of colors, represented by sRGB triplets, and by respecting a certain number of rules defined from a specification. Such a specification typically includes criteria of both cognition and communication, these criteria being able to be in contradiction.

[0027] The communication criteria concern the search for an aesthetic aspect, compliance with a graphic charter or even conformity to a specific evocative application, for example the desired color of representation of animals for hunting images.

[0028] The cognition criteria aim, for example, to facilitate the interpretation of the image according to the type of scene, to improve the perception of the contrast of the image, to increase the salience on the type of objects of interest or to facilitate the interpretation of the image for people with visual impairments, typically for color blind people.

[0029] In all cases, the criteria of cognition are distinct from one person to another.

[0030] However, existing solutions allow for the design of linear color palettes in absolute luminance, but they are not linear in terms of perception with regard to the cognition criteria of a specific person. It follows that certain details of a colorized infrared image are not necessarily visible to a given person, whereas they may be visible to another person.

[0031] The technical problem of the invention is therefore to obtain a method for colorizing an infrared image with a linear color palette as a function of the

[0032]

[0033]

[0034]

[0035]

[0036]

[0037]

[0038]

[0039]

[0040] a person's perception, so that a person can simply modify the color palette for a specific application and / or for personal perception. Statement of the invention To address this technical problem, the invention proposes to use a color palette defined in the CIE CAM02-UCS space by a parametric trajectory. This uniform CIE CAM02-UCS color space is defined in the article R. Luo, G. Cui and C. Li, “Uniform color spaces based on CIECAM02 color appearance model” Color Research and Application, 2006. In this Cartesian space of axes j*, a* and b*, j* corresponds to the perceived luminance, a* corresponds to the blue to yellow color axis and b* corresponds to the green to red color axis. Chrominance is defined by the following equation: [Math.sl] - / >2 , *2 c ' - " + b The hue is defined by the following equation: [Math.s2] h* = atanQ^ j More specifically, the invention stems from an observation that, to achieve effective linearity in terms of perception, the parameter of this trajectory must be the polar angle vector. According to this same observation, the trajectory projected in the hue and chrominance plane must also follow one or two ellipses while the dynamics of the perceived luminance with respect to the grayscale values ​​must follow one or two straight lines. With the invention, a user can use predefined mathematical definitions of this parametric trajectory for different applications and obtain a color palette that is linear in terms of perception with respect to generic cognition criteria. In addition, the user can also adjust certain parameters of this parametric trajectory to adapt a generic cognition criterion to his own feelings. Thus, the invention relates to a method for colorizing an infrared image comprising the following steps: - acquisition of an infrared image obtained from an infrared sensor comprising a set of elementary detectors, said infrared image comprising, for each elementary detector, a pixel whose value is coded in gray levels corresponding to a value representing infrared radiation received by said elementary detector; and - generation of a colorized image by associating with each pixel value coded in grayscale a pixel value coded in a color space using a color palette.

[0041] The invention is characterized in that said color palette is defined in the CIE CAM02-UCS space by a palette generation module using a parametric trajectory in which a parameter is constituted by a polar angle vector of the parametric trajectory, the latter being defined so that: - the parametric trajectory projected into a hue and chrominance plane follows one or two ellipses; and - a dynamic of the perceived luminance in relation to the gray level values ​​follows one or two straight lines.

[0042] Preferably, the parametric trajectory follows the following relationship:

[0043] [Math.s3] = ad](cos$-1) . b'( / ) = b^sin^

[0044] with a first half axis (aeU) corresponding to a chrominance at the apogee (cc*) divided by two, and with a second half axis (beU) defined as a function of the first half axis (aeu) with:

[0045] [Math.s4]

[0046] if an eccentricity (e) of the ellipse is positive; and

[0047] [Math.s5]

[0048] if eccentricity (e) of the ellipse is negative.

[0049] With these precise construction parameters, the invention makes it possible to obtain a color palette designed linearly in the CIE CAM02-UCS space in order to guarantee good linearity of luminance perception.

[0050] The color palette thus defined can then be modified by a user according to his own expectations. To do this, the method can also include the following steps: - displaying said colorized image on a screen visible to a user; and - recovery, from a human-machine interface, of the colorization modifications expected by the user.

[0051] Preferably, the human-machine interface makes it possible to modify a hue at the peak, a chrominance at the peak, a dynamic of the polar angle vector, an eccentricity of said at least one ellipse projected in the hue plane and chrominance, and / or a dynamic of perceived luminance. These parameters make it possible to modify a color palette with notions that are intelligible to a person skilled in the art of improving image chromatography.

[0052] For the purposes of the invention, the peak of the hue or chrominance refers to the point for which the polar angle corresponds to jt. In the case where the ellipticity is positive, this peak corresponds to a maximum of chrominance.

[0053] Preferably, the dynamics of the polar angle vector is limited between 0 and 2ir or 2ir and 0.

[0054] Furthermore, to simplify the selection of a color palette for specific applications, said human-machine interface can also allow the selection of predefined palettes.

[0055] For example, to evoke the idea of ​​thermal imaging, while avoiding unattractive purple hues and overly aggressive white or yellow hues, a first predefined palette preferably has a peak hue of between 20 and 30°, a peak chrominance of between 30 and 40, a dynamic range of the polar angle vector of between 0 and 1.8ji, an eccentricity of an ellipse projected into the hue and chrominance plane of between -0.35 and -0.25, and a dynamic range of the perceived luminance of between 5 and 100. This first predefined palette avoids pure white and overly bright yellow for the representation of warm tones, in order to limit the effect of aggression and visual fatigue.

[0056] For night vision applications, the green monochromatic representation of a night image takes as a reference the phosphor scintillator used in light intensifier tubes, which has a spectral re-emission length centered on green. The use of green has become iconic and evocative to the general public of night imaging, since it is carried in particular by military propaganda, through cinema and video games. The use of green also has advantages in terms of sensitivity to light, since the spectral sensitivity curve is maximum around 555 nanometers, at least in good lighting conditions.

[0057] In this type of night vision application, the display is generally set to very low brightness. This makes it possible, on the one hand, to reduce the user's vision adaptation time when he stops viewing the colorized infrared image display screen to directly observe his environment; and on the other hand, it makes it possible to reduce the light reflected on the face, for stealth reasons.

[0058] However, in scotopic vision, the spectral sensitivity curve adapts to darkness, involving more rods rather than cones; this is then scotopic vision whose sensitivity curve is centered around 507 nanometers.

[0059] The second predefined palette is designed to facilitate viewing of an infrared image in dark conditions, while limiting visual fatigue in the event of prolonged viewing of a video stream. To this end, the second predefined palette has a peak hue of between -170 and -160°, a peak chrominance of between 10 and 20, a dynamic range of the polar angle vector of between 0 and 2ir, an eccentricity of an ellipse projected into the hue and chrominance plane of between 0.2 and 0.4, and a dynamic range of the perceived luminance of between 100 and 20.

[0060] With this second predefined palette, the image appears soft, round, while remaining sharp and prominent on the objects of interest. The background is displayed in light, especially for the sky, which reduces the efforts both on the opening of the pupil and on the interpretation, in particular by familiarity with daytime scenes where the sky is clear. The aesthetic aspect of the colorized image gives a glossy paper print appearance, considered rather pleasant.

[0061] A third predefined palette is constituted by a threshold palette, that is to say a palette which is constructed by concatenation of two palettes, with a gray level threshold from which a transition between the two palettes is made. Preferably, a continuity of colors is ensured at the level of said threshold. These two palettes make it possible to separate two families / classes of pixels by coloring them differently, and by maximizing the dynamic range of luminance perceived on each class. These two families / classes of pixels can be distinguished by variations in distinct hues or chrominances.

[0062] To bring out the warm areas of the ambient temperature of a scene, by maximizing the details on the two areas, the third predefined palette preferably has a first portion with a hue at the peak between -110 and -100°, a chrominance at the peak between 15 and 20, a dynamic range of the polar angle vector between 2ir and 0, an eccentricity of a first ellipse projected in the hue and chrominance plane between 0.5 and 0.7, and a dynamic range of the perceived luminance between 100 and 5; then a second portion with a peak hue between 25 and 40°, a peak chrominance between 30 and 40, a dynamic range of the polar angle vector between 0 and 1.8ir, an eccentricity of a second ellipse projected into the hue and chrominance plane between 0.3 and 0.4, and a dynamic range of perceived luminance between 5 and 99, the threshold between the two portions being set between 45 and 55% of the grayscale values.

[0063] In this third predefined palette, the representation of the object of interest is made by a palette close to the first predefined palette, with the difference that the hue dynamics are slightly reduced so as not to overload the image too much.

[0064] Thus, the object of interest is colored normally, which facilitates the interpretation of the nature of the image. The perceived luminance has a full and increasing dynamic range to ensure linearity of color evolution for the user.

[0065] The background is also represented by means of an elliptical palette, with a full dynamic decreasing perceived luminance, and a slate blue apogee. This apogee color is chosen at low chrominance because it is desired to pay less attention to the background and more attention to the object of interest. Similarly, with this third predefined palette, it is desired to limit the variation in hue via a trajectory with high ellipticity. These choices also have the consequence of reducing the sensitivity to small variations in intensity, partly limiting the perception of noise, the signal-to-noise ratio being generally lower on the ambient background than on the warm object of interest.

[0066] The decreasing perceived luminance makes it possible to lighten the background to improve the saliency on the object of interest. This saliency is supported by a black outline at the threshold, and by the chrominance contrast between the two classes of pixels. To do this, the pixels around the threshold are colorized with dark colors with low chrominance, typically colors close to black. These pixels are generally at the border between the background and the object of interest.

[0067] Finally, the use of blue for the representation of the background is justified on the one hand because it is a shade that evokes a cold temperature, and on the other hand because it is a color that evokes the sky, which is a reassuring representation for humans. For this same reason, the use of blue creates an illusion of depth by recalling the horizon, at least it is a technique used by the Impressionists to produce this effect.

[0068] In a fourth predefined palette, it is sought to define a palette which includes a grayscale section with full perceived luminance dynamics and, from a warm threshold, a nearby bright colorized section with red tints.

[0069] For hunting applications, the hot spots are generally animals, which are sought to be highlighted in the image. The environment is also very important in such an application. It is generally made up of complex elements such as vegetation, hence the importance of representing it in full gray dynamics. Other applications may need to alert the user, for example the AD AS application, for "Advanced Driver Assistance Systems" in the English literature, consisting of alerting a driver to the presence of a pedestrian, for braking assistance.

[0070] To do this, the fourth predefined palette preferably presents a first portion with zero chrominance and a dynamic range of the perceived luminance between 0 and 100; then a second portion with a hue at the peak between 30 and 35°, a chrominance at the peak between 30 and 40, a dynamic range of the polar angle vector between 2ir and 0, an eccentricity of an ellipse projected in the hue and chrominance plane between 0.8 and 0.9, and a dynamic range of the perceived luminance between 100 and 30, the threshold between the two portions being fixed between 75 and 85% of the grayscale values. Summary description of the figures

[0071] The manner of carrying out the invention as well as the advantages which result therefrom will emerge clearly from the following embodiments, given for informational but non-limiting purposes, supported by figures 1 to 7 in which:

[0072] [Fig.l] is a schematic representation of a method for colorizing an infrared image according to one embodiment of the invention;

[0073] [Fig.2] is a schematic representation of a parametric generation of a color palette according to the method of [Fig.l] with an ellipse projected into a hue and chrominance plane;

[0074] [Fig.3] illustrates three variations of a parametric generation of a color palette according to the method of [Fig.l] with an ellipse projected into the hue and chrominance plane as a function of the variation of the eccentricity of the ellipse;

[0075] [Fig.4] illustrates the dynamics of the perceived luminance and of an ellipse projected in the hue and chrominance plane for a first predefined palette of the method of [Fig.l];

[0076] [Fig.5] illustrates the dynamics of perceived luminance and of an ellipse projected into the hue and chrominance plane for a second predefined palette of the method of [Fig.l];

[0077] [Fig.6] illustrates the dynamics of the perceived luminance and two ellipses projected in the hue and chrominance plane for a third predefined palette of the method of [Fig.l]; and

[0078] [Fig.7] illustrates the dynamics of perceived luminance and an ellipse projected into the hue and chrominance plane for a fourth predefined palette of the method of [Fig.l], Detailed description of the invention

[0079] [Fig.l] illustrates a method for colorizing an infrared image 11 obtained from an infrared sensor 10 comprising a set of elementary detectors. Conventionally, the elementary detectors consist of microbolometers arranged in matrix form.

[0080] For each row or for each column of microbolometers, a reading circuit makes it possible to measure a quantity of infrared radiation captured by each elementary detector, thus forming the pixels of an infrared image.

[0081] The values ​​from each pixel of the infrared image can be coded in

[0082]

[0083]

[0084]

[0085]

[0086]

[0087]

[0088]

[0089] ,2 a" + b NG grayscale on 8, 11 or 14 bits without changing the invention. In addition, pre-processing can be carried out on the infrared image 11 before implementing the colorization method of the invention. More specifically, the invention proposes using a color palette 12 to transform the infrared image 11 coded in gray levels NG into a colorized image 13, each pixel of which is coded by a color system, for example according to the sRGB standard. According to the invention, the color palette 12 is defined in the CIE CAM02-UCS space, defining each color by six characteristics: brightness, absolute luminance, colorization, chrominance c*, saturation and hue h*. These six characteristics are interdependent and can be coded by a triplet of three values: j*, a*, and b*, with the parameter]* corresponding to the perceived luminance, a* corresponding to the blue to yellow color axis and b* corresponding to the green to red color axis. In this CIE CAM02 -UCS space, the chrominance c* is defined by the following equation: [Math.sô] c* = The hue h* is defined by the following equation: [Math.s7] h* = atan^^ More specifically, the invention proposes to generate the color palette 12 in the CIE CAM02-UCS space by a palette generation module 20 using a parametric trajectory in which the parameter is a polar angle vector ¢. This parametric definition is constrained so that the trajectory projected in the hue and chrominance plane (O, ac*, bc*) follows one or two ellipses; and the dynamics of the perceived luminance j* with respect to the grayscale values ​​NG follows one or two straight lines. [Fig.2] illustrates a representation of an ellipse in the hue and chrominance plane (O, ac*, bc*) of a parametric generation of a color palette 12. In this example, the ellipse is formulated according to a parametric trajectory, whose parameter is the polar angle vector 0 = [0,.., (f)max] of length 2n, with n corresponding to the depth of the digitized pixel value. Thus, the different gray level values ​​NG are between 0 and 2nA. Preferably, the depth 0mar= l.Sir. Preferably, the dynamics of the polar angle vector 0 is limited between 0 and 2ir or 2ir and 0.

[0090] In the example of [Fig.2], the ellipse does not complete a complete rotation, that is to say it is not closed, to avoid ending the palette on a pure white which is considered aggressive.

[0091] The ellipse is defined according to the following equation:

[0092] [Math.s8] = 1) . è'(0) = b eil 5m0

[0093] in which the ellipse has a first half axis aeu corresponding to a chrominance at the apogee cc* divided by two, in the frame aligned with the axes of the ellipse (O, a', b'), and the second half axis beu depends on the corresponding first half axis aeu and the eccentricity e of the ellipse.

[0094] More precisely, the second half axis b,^c deduced from the desired eccentricity e, with the following equation, if the eccentricity is positive:

[0095] [Math.s9]

[0096] When the eccentricity is negative, the second half axis follows the following equation:

[0097] [Math.slO]

[0098] When the eccentricity is positive, the first half axis aeU is larger than the second half axis beu while the first half axis aeu is smaller than the second half axis beu when the eccentricity is negative.

[0099] The impact of varying the eccentricity e is illustrated in [Fig.3] in which a parametric trajectory with a low eccentricity allows one to take advantage of the purplish and orange nuances, while a parametric trajectory with a high eccentricity avoids these nuances, at the risk of giving the image a monochrome appearance.

[0100] For cosmetic reasons and to facilitate the perception of details via good dynamics of purplish / orange hues, it is preferable to select a low eccentricity, typically an eccentricity between -0.2 and -0.4, for example e=-0.3.

[0101] Whatever the value of the chosen eccentricity, it is possible to make a change of reference to write the ellipse in the reference (O, a*, b*) of [Fig.2] by simple azimuthal rotation of angle hc*, that is to say by the desired apogee hue, according to the following equation:

[0102] [Math.sll] “( $ ) = a ( (J> ) cos / l - b ( <J) ) sin / L = a' (¢) sin / ?c + b () cos / l

[0103] The apogee hue angle is written / zc*=atan2(Z?c* ac*) and the perceived luminance curve j*(0) is simply a straight line whose dynamic range limits can be set. The values ​​of apogee hue hc*, apogee chrominance cc*, dynamic range of the polar angle vector ¢, eccentricity e, and / or a dynamic range of the perceived luminance j*, can be selected according to the type of color palette sought. For example, it is possible to increase the hue dynamic range by reducing the eccentricity e of the ellipse.

[0104] With these precise parameters for defining the color palette 12, it is possible to obtain a table associating each gray level value NG with a triplet in the sRGB space. For example, this conversion between the CIE CAM02-UCS space and the sRGB space can be carried out by the “colorspacious” module, coded in the Python computer language. This table can be called LUT for “Look Up Table” in the English literature.

[0105] In this color palette 12, each grayscale value NG is therefore associated with a color triplet in the sRGB space so that the infrared image 11 can be transformed into a colorized image 13. The colorized image 13 can then be displayed on a screen 14 so that a user can effectively detect the salient elements of a scene captured by an infrared sensor 10.

[0106] To detect specific elements in various applications, the user can use a human-machine interface 15 to select predefined color palettes.

[0107] In addition, the human-machine interface 15 may also allow the user to adjust the construction values ​​of the color palette 12.

[0108] Indeed, even if the color palette 12 is defined with values ​​of hue at the peak hc*, chrominance at the peak cc*, dynamics of the polar angle vector ¢, eccentricity e, and / or a dynamics of the perceived luminance j*, the user can possibly modify these values ​​through the human-machine interface 15. To do this, the user can request an increment or a decrease of each value in order to observe the evolution of these variations directly on the screen 14 for viewing the colorized image 13.

[0109] With respect to the predefined color palettes, four palettes can be offered to the user via the human-machine interface 15: Lifeinred™ Color, Lifeinred™ Serenity, Lifeinred™ Contrast, and Lifeinred™ Tracker.

[0110] For example, the predefined palette Lifeinred™ Color has a peak hue hc* between 20 and 30°, a peak chrominance ec* between 30 and 40, a dynamics of the polar angle vector between 0 and 1.8rr, an eccentricity e of an ellipse projected in the hue and chrominance plane (O, a*,b*) between -0.35 and -0.25, and a dynamic range of the perceived luminance j* between 5 and 100.

[0111] As illustrated in [Fig.4], this predefined Lifeinred™ Color palette has a straight and increasing perceived luminance dynamic j* and an ellipse in the hue and chrominance plane (O, a*, b*). Point PI corresponds to a coral color, point P2 corresponds to a vanilla or peach color and point P3 corresponds to a wine color.

[0112] The Lifeinred™ Serenity predefined palette has a peak hue hc* between -170 and -160°, a peak chrominance cc* between 10 and 20, a polar angle vector dynamics between 0 and 2ir, an eccentricity e of an ellipse projected in the hue and chrominance plane (O, a*,b*) between 0.2 and 0.4, and a dynamic range of the perceived luminance j* between 100 and 20.

[0113] As illustrated in [Fig.5], this predefined Lifeinred™ Serenity palette has a straight and decreasing perceived luminance dynamics j* with a reduced dynamics Dr and an ellipse in the hue and chrominance plane (O, a*,b*). Point P6 corresponds to a sea green color, while point P5 corresponds to a slate gray color and point P7 corresponds to the teal blue color.

[0114] The predefined Lifeinred™ Contrast palette has a first portion with a peak hue hc* between -110 and -100°, a peak chrominance ec* between 15 and 20, a dynamic range of the polar angle vector between 2ir and 0, an eccentricity e of a first ellipse projected in the plane (O, a*,b*) between 0.5 and 0.7, and a dynamic range of the perceived luminance j* between 100 and 5.

[0115] The predefined Lifeinred™ Contrast palette also features a second portion with a peak hue hc* between 25 and 40°, a peak chrominance cc* between 30 and 40, a polar angle vector dynamics between 0 and 1.8ir, an eccentricity e of a second ellipse projected in the plane (O, a*,b*) between 0.3 and 0.4, and a dynamic range of the perceived luminance j* between 5 and 100. The threshold SI between the two portions is fixed between 45 and 55% of the values ​​in gray levels NG, for example 50%.

[0116] As illustrated in [Fig.6], this predefined Lifeinred™ Contrast palette includes a dynamic of the perceived luminance j* right and decreasing up to the threshold SI then a dynamic of the perceived luminance j* right and increasing. Two ellipses are formed in the plane (O, a*,b*). The point PI corresponds to a coral color, also used for the predefined Lifeinred™ Color palette, while the point P5 corresponds to the color slate blue.

[0117] The predefined Lifeinred™ Tracker palette presents a first portion with zero chrominance and a perceived luminance dynamic range j* between 0 and 100; then a second portion with a peak hue hc* between 30 and 35°, a chrominance at the peak cc* between 30 and 40, a dynamic range of the polar angle vector 0 between 2ir and 0, an eccentricity e of an ellipse projected in the plane (O, a*,b*) between 0.8 and 0.9, and a dynamic range of the perceived luminance j* between 100 and 30. The threshold S2 between the two portions being fixed between 75 and 85% of the values ​​in gray levels NG, for example 80%.

[0118] As illustrated in [Fig.7], this predefined Lifeinred™ Tracker palette includes a right and increasing perceived luminance dynamics j* up to the threshold S2 and then a right and decreasing perceived luminance dynamics j*. An ellipse is formed in the ab plane, in which the point P8 corresponds to a brick red color.

[0119] With these different predefined color palettes and the human-machine interface 15 allowing the user to modify the generation parameters of the color palette 12, the invention allows a user to obtain a linear rendering according to his personal perception. The invention thus makes it possible to observe more distinctly the objects of interest captured by an infrared detector.

Claims

Claims

1.

2.

3.

4. Method for coloring an infrared image (11) comprising the following steps: - acquisition of an infrared image (11) obtained from an infrared sensor (10) comprising a set of elementary detectors, said infrared image (11) comprising, for each elementary detector, a pixel whose value is coded in gray levels (NG) corresponding to a value representing infrared radiation received by said elementary detector; and - generation of a colorized image (13) by associating with each pixel value coded in gray levels (NG) a pixel value coded in a color space by means of a color palette (12); characterized in that said color palette (12) is defined in the CIE CAM02-UCS space by a palette generation module (20) using a parametric trajectory in which a parameter is a polar angle vector (¢) of the parametric trajectory, the latter being defined so that: - the parametric trajectory projected into a hue and chrominance plane (O, a*, b*) follows one or two ellipses; and - a dynamic of the perceived luminance (j*) in relation to the gray level values (NG) follows one or two straight lines. A method of colorizing an infrared image according to claim 1, wherein the parametric trajectory follows the following relationship: = &e^coS(p - 1) with a first half axis (aeù) corresponding to a chrominance at the apogee (cc*) divided by two, and a second half axis (ben) defined as a function of the first half axis (aeU) with: — a^l - e 2 if an eccentricity (e) of the ellipse is positive; and — -244; if eccentricity (e) of the ellipse is negative. Method for colorizing an infrared image according to claim 1 or 2, wherein the method also comprises the following steps: - displaying said colorized image (13) on a screen visible (14) by a user; and - recovery, from a human-machine interface (15), of the colorization modifications expected by the user. Method for coloring an infrared image according to claim 3, in which the human-machine interface (15) makes it possible to modify a hue at the peak (hc*), a chrominance at the peak (cc*), a dynamic of the polar angle vector (¢), an eccentricity (e) of the at least one ellipse projected in the hue and chrominance plane (0, a*, b*), and / or a dynamic of the perceived luminance (j*).

5. A method of coloring an infrared image according to one of claims 1 to 4, wherein the dynamics of the polar angle vector (¢) is limited between 0 and 2ir or 2ir and 0.

6. Method for coloring an infrared image according to one of claims 1 to 5, in which said human-machine interface (15) makes it possible to select predefined palettes.

7. A method for colorizing an infrared image according to claim 6, wherein a first predefined palette (Lifeinred™ Color) has a peak hue (hc*) of between 20 and 30°, a peak chrominance (cc*) of between 30 and 40, a dynamic range of the polar angle vector (¢) of between 0 and 1.8ir, an eccentricity (e) of an ellipse projected into the hue and chrominance plane (O, a*,b*) of between -0.35 and -0.25, and a dynamic range of the perceived luminance (j*) of between 5 and 100.

8. A method of colorizing an infrared image according to claim 6 or 7, wherein a second predefined palette (Lifeinred™ Serenity) has a peak hue (hc*) of between -170 and -160°, a peak chrominance (cc*) of between 10 and 20, a dynamic range of the polar angle vector (¢) of between 0 and 2ir, an eccentricity (e) of an ellipse projected into the hue and chrominance plane (O, a*,b*) of between 0.2 and 0.4, and a dynamic range of the perceived luminance (j*) of between 100 and 20.

9. A method of colorizing an infrared image according to one of claims 6 to 8, wherein a third predefined palette (Lifeinred™ Contrast) has a first portion with a peak hue (hc*) between -110 and -100°, a peak chrominance (cc*) between 15 and 20, a dynamic range of the polar angle vector (¢) between 2ir and 0, an eccentricity (e) of a first ellipse projected into the hue and chrominance plane (O, a*,b*) between 0.5 and 0.7, and a dynamic range of the perceived luminance (j*) between 100 and 5; then a second portion with a hue at the peak (hc*) between 25 and 40°, a chrominance at the peak (cc*) between 30 and 40, a dynamic of the polar angle vector (¢) between between 0 and 1.8ji, an eccentricity (e) of a second ellipse projected in the hue and chrominance plane (O, a*,b*) between 0.3 and 0.4, and a dynamic range of the perceived luminance (j*) between 5 and 100, the threshold (SI) between the two portions being fixed between 45 and 55% of the grayscale values (NG).

10. Method for colorizing an infrared image according to one of claims 6 to 9, in which a fourth predefined palette (Lifeinred™ Tracker) has a first portion with zero chrominance and a dynamic range of the perceived luminance (j*) between 0 and 100; then a second portion with a peak hue (hc*) between 30 and 35°, a peak chrominance (cc*) between 30 and 40, a dynamic range of the polar angle vector (¢) between 2ir and 0, an eccentricity (e) of an ellipse projected in the hue and chrominance plane (O, a*,b*) between 0.8 and 0.9, and a dynamic range of the perceived luminance (j*) between 100 and 30, the threshold (S2) between the two portions being set between 75 and 85% of the grayscale values (NG).