Converting an image to an output image having a higher dynamic range

Abstract

The image processor converts an input image having a first dynamic range into an output image having a second dynamic range higher than the first dynamic range. The image processor converts input luminance values ​​(L sdr ) to the output luminance value (L hdr The mapping unit (10) is configured to apply a continuously increasing function that maps a first range of values ​​to a second range of values, whereby the highest value in the first range of values ​​is mapped to the upper limit of the second range of values ​​by the continuously increasing function. The mapping unit (10) converts the peak luminance (Y peak ) and configured to adjust the upper bound depending on the input image. Methods for converting an input image to an output image and other image processing devices are also provided.

Description

[Technical field]

[0001] The present invention relates to the field of image processing.

[0002] More particularly, the present invention relates to an image processing apparatus and method for converting an input image to an output image. [Background technology]

[0003] Image processing devices have been proposed for converting an input image having a first dynamic range (e.g., "standard dynamic range" or SDR) to an output image having a second dynamic range (e.g., "high dynamic range" or HDR) that is higher than the first dynamic range. Such conversion is commonly referred to as "tone expansion."

[0004] In such an image processing device, a mapping unit is provided for converting an input luminance value associated with a pixel of the input image to an output luminance value associated with a corresponding pixel in the output image.

[0005] The mapping unit is configured to apply a continuously increasing function that maps the first numerical range to a second numerical range, whereby a maximum value in the first numerical range is mapped to an upper limit of the second numerical range by the continuously increasing function.

[0006] This function is generally designed to take advantage of the second dynamic range extension, with the upper limit conventionally corresponding to the maximum luminance achievable by the HDR display from which the output image is intended to be obtained. Summary of the Invention [Means for solving the problem]

[0007] In this regard, the present invention provides an image processing device for converting an input image having a first dynamic range into an output image having a second dynamic range higher than the first dynamic range, the image processing device comprising a mapping unit for converting input luminance values ​​associated with a pixel of the input image into output luminance values ​​associated with a corresponding pixel in the output image, the mapping unit being configured to apply a continuously increasing function that maps the first numerical range to the second numerical range, a highest value in the first numerical range being mapped by the continuously increasing function to an upper limit of the second numerical range, and the mapping unit being configured to adjust the upper limit depending on the peak luminance.

[0008] Due to the adjustable upper limit, the mapping function and the resulting peak luminance can be adapted (either manually, as will be explained later, or automatically) depending on the image being processed. In particular, the inventors have realized that mapping the luminance over the entire available dynamic range for the output image does not always produce the most pleasing results.

[0009] According to a possible embodiment, the mapping unit may comprise a module for identifying the peak luminance as a function of the pixel values ​​of the input image. The upper limit and the mapping function may therefore be automatically adjusted depending on the content of the image being processed.

[0010] The peak luminance identifying module is, for example, configured to identify the peak luminance based on luminance values ​​respectively associated with pixels of the input image.

[0011] According to the embodiment described below, the module for identifying peak values ​​is configured to evaluate the brightness values ​​of the input image and to identify a peak luminance as a function of said values ​​(eg as a decreasing function).

[0012] The mapping unit may also be configured to vary at least one parameter of the continuously increasing function depending on the peak luminance, eg such that the curvature of the applied continuously increasing function increases as a function of the peak value.

[0013] According to an implementation described below, the mapping unit includes a module for determining an exponent based on a peak luminance, and a module for exponentiating with the exponent.

[0014] In this implementation, the module for identifying the exponent may then be configured such that the exponent increases as a function of the peak luminance.

[0015] However, according to a possible variant, other calculation modules of the mapping unit (eg a static mapping module) may perform processing that can vary depending on the peak luminance.

[0016] The mapping unit may include a (calculation) module configured to convert the input luminance value to an intermediate luminance value, for example by applying a mapping function depending on the peak luminance, to which the above-mentioned power module may be applied (i.e. the power module raises the intermediate luminance value to the above-mentioned exponent).

[0017] The present invention also provides a method for converting an input image having a first dynamic range into an output image having a second dynamic range higher than the first dynamic range, comprising the step of converting an input luminance value associated with a pixel of the input image into an output luminance value associated with a corresponding pixel in the output image, the converting step comprising the sub-step of applying a continuously increasing function that maps the first numerical range to a second numerical range, whereby a maximum value of the first numerical range is mapped by the continuously increasing function to an upper limit of the second numerical range, the upper limit being adjustable depending on the peak luminance.

[0018] Such a method may further include identifying a peak luminance as a function of pixel values ​​of the input image.

[0019] This step of identifying the peak luminance may include the sub-step of evaluating the brightness values ​​of the input image and / or the sub-step of identifying the peak luminance as a function of said values.

[0020] As previously mentioned, the parameters of the continuously increasing function may be variable depending on the peak luminance.

[0021] Finally, the present invention provides an image processing device for converting an input image having a first dynamic range into an output image having a second dynamic range lower than the first dynamic range, the image processing device comprising a mapping unit for converting input luminance values ​​associated with a pixel of the input image into output luminance values ​​associated with a corresponding pixel in the output image, the mapping unit being configured to apply a continuously increasing function that maps the first numerical range into the second numerical range, an upper limit of the first numerical range being mapped by the continuously increasing function to a highest value in the second numerical range, the mapping unit being configured such that the upper limit is adjustable depending on the peak luminance.

[0022] Other possible features of these devices and methods will become apparent from the following description of possible embodiments of the invention.

[0023] The following description of the accompanying drawings will make it clear what the invention is and how it can be implemented. The invention is not limited to the embodiments shown in the drawings. Therefore, when features specified in the claims are accompanied by reference signs, it should be understood that such signs are included only for the purpose of making the claims more understandable, and do not limit the scope of the claims in any way. [Brief description of the drawings]

[0024] [Figure 1]1 shows an example of an image processing device according to the present invention. [Diagram 2] 2 shows a possible embodiment of a mapping unit usable in the image processing device of FIG. 1; [Diagram 3] 3 shows curves representing examples of continuously increasing functions used by the mapping unit of FIG. 2; [Figure 4] 3 shows a possible embodiment of a peak luminance providing module that can be used in the mapping unit of FIG. 2; [Diagram 5] 2 illustrates an example of an image processing device that can be used with the image processing device of FIG. 1 according to another aspect of the present invention. [Figure 6] 6 shows a possible embodiment of a mapping unit usable in the image processing device of FIG. 5; DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0025] FIG. 1 shows an example of an image processing apparatus according to the present invention.

[0026] The image processing device 1 may in fact be implemented by an electronic device including a processor and a memory storing program code instructions adapted to perform the operations and functions of the modules and units described below when the relevant program code instructions are executed by the processor.

[0027] As will be apparent from the following description, the image processing device 1 has a first dynamic range Δ sdr (e.g., standard dynamic range, or SDR) sdr The second dynamic range Δ hdr (e.g., high dynamic range, or HDR) output image I hdr The second dynamic range Δ hdr is the first dynamic range Δ sdr (strictly) higher.

[0028] First Dynamic Range Δ sdrInput image I sdr The first dynamic range Δ sdr Higher second dynamic range Δ hdr Output image I hdr Such a process of converting to is commonly called "tone expansion."

[0029] Input image I sdr is the multiple component values ​​R for each pixel p. p , G p , B p Then, the input image I sdr It is represented using a group of pixels (generally a matrix of pixels).

[0030] In this example, the input image I sdr are the three color components R for each pixel p. p , G p , B p (i.e., the red component R p , green component G p , and the blue component B p However, for example, the luminance component Y p and the two chromaticity components U p , V p etc., using input image I sdr Other representations for may also be used.

[0031] As can be seen in FIG. 1, the image processing device 1 receives an input image I sdr , i.e., the input image I sdr The component value R p , G p , B p In practice, the input image I sdr The component value R p , G p , B p can be received from other modules of the electronic device or from external electronic device (through a communication circuit cooperating with the input module 2).

[0032] The image processing device 1 also receives the input luminance value L sdrthe component value R (of the associated pixel p) p , G p , B p The image generation module 4 is configured to generate the image based on the

[0033] The image creation module 4 receives the input luminance value L sdr component value R p , G p , B p In this case, several steps are required to generate the - Inverse Optical-Electrical Transfer Function (i.e., OETF -1 ) application, - Conversion from ITU-R BT.709 standard representation to ITU-R BT.2020 standard representation, - Luminance calculation from ITU-R BT.2020 standard representation can be executed.

[0034] The input luminance values ​​L generated by the image creation module 4 by the step of applying the inverse photoelectric transfer function sdr is the input image I sdr Represents the linear luminance component of

[0035] The image processing device receives the input image I sdr The input luminance value L associated with any pixel of sdr The output image I hdr The output luminance value L associated with the corresponding pixel p' of hdr The mapping unit 10 (described further below) is designed to convert

[0036] As described below, the mapping unit 10 calculates a continuously increasing function f that maps a first range of values ​​[x0, x1] to a second range of values ​​[y0, y1]. Ypeak whereby the highest value x1 in the first numerical range is multiplied by a continuously increasing function f Ypeak The upper limit y1 of the second range of values ​​is selected by a user of the image processing device or by the input image I sdr The peak luminance Y can be determined based onpeak It is adjustable depending on (and relative to)

[0037] Function f Ypeak is an increasing function if, for any pair (z1, z2) in the first numerical range [x0, x1], z1≦z2, f Ypeak (z1)≦f Ypeak This means that (z2).

[0038] As will be explained further below, the continuously increasing function f Ypeak is the adjustable peak value Y peak It varies with f Ypeak It is written as follows.

[0039] Continuously increasing function f Ypeak is convex in the first range of values. (A continuously increasing function f Ypeak In particular, the peak luminance Y peak is the maximum brightness Y max Such a function provides a wider tonal extension in the highlights (compared to the shadows and midtones) and a (wider) second dynamic range Δ hdr However, this effect is at the expense of a lower peak luminance Y peak The value of is sparse, as will be explained later.

[0040] In this case, the continuously increasing function f Ypeak is the input luminance value L sdr (included in the first numerical range [x0,x1] and associated with a particular pixel p) as the output luminance value L hdr (The second numerical range [y0,y1] is included in the output image I hdr , which is associated with the corresponding pixel p' in

[0041] In this case, the input luminance value L sdr are normalized (i.e. they represent a ratio to the maximum possible input luminance value), so that x0=0 and x1=1.

[0042] A possible embodiment of the mapping unit 10 is described below with reference to FIG.

[0043] The image processing device 1 outputs an output image I hdr For each pixel p' in p’ , G' p’ , B' p’ Let p′ be the output luminance value L generated by the mapping unit 10 for the associated pixel p′. hdr and the input image I sdr The component value R for the corresponding pixel p in p , G p , B p The image generation module 6 is configured to identify the object based on

[0044] Component value R' p’ , G' p’ , B' p’ The output brightness value L hdr and component value R p , G p , B p In order to identify based on, the image generation module 6 may, for example, - Output luminance value L hdr Using the component value R p , G p , B p (The scaled component value is the output luminance value L hdr ), - performing color correction (e.g. by adjusting saturation), - the possible step of applying an optical-electrical transfer function (i.e. OETF); Execute.

[0045] In this example, the component value R' p’ , G' p’ , B' p’ is the output image I hdr For each pixel p', the three color components R' p’ , G' p’ , B' p’ (i.e., the red component R' p, green component G' p’ , and the blue component B' p’ ) However, the output image I hdr Other representations may be used for, for example, the luminance component Y' for each pixel p'. p’ (In this case, the output luminance value L generated by the mapping unit 10 for the pixel p′ concerned hdr ) and the two chromaticity components U' p’ , V' p’ is used.

[0046] The image processing device 1 calculates the component value R' p’ , G' p’ , B' p’ An output module 8 for outputting an output image I on a screen of said electronic device, for example by a display module of said electronic device. hdr is used to display the image or, in a modified form, is transmitted to an external electronic device (using the electronic device's communication circuitry).

[0047] In other words, the electronic device in which the image processing device 1 is implemented outputs an output image I hdr However, in a variant, the electronic device may also be a processing device without a display, possibly displaying the output image I hdr The component value R' represents p’ , G' p’ , B' p’ to an external electronic device (output image I hdr and a communications circuit for transmitting the signal to a display device (which may include a screen suitable for displaying the signal).

[0048] FIG. 2 shows a possible embodiment of the mapping unit 10 .

[0049] The mapping unit 10 includes a peak luminance providing module 11 , a static mapping module 12 , a central tendency assessment module 14 , a central tendency correction module 15 , an index identification module 16 , and an adaptive mapping module 18 .

[0050] The peak luminance providing module 11 provides a peak luminance Y peak This peak luminance Y peak is a first predetermined value Y min and a second given value Y max It is possible to adjust between

[0051] A (second) given value Y max Therefore, the adjustable peak brightness Y peak This is the maximum allowable value of the hdr may correspond to the maximum luminance obtainable by the display on which it is to be displayed. For current HDR displays, 1000 nits (i.e., 1000 cd / m 2 ) maximum luminance Y max is used.

[0052] Peak Luminance Y peak Since is adjustable, this of course corresponds to the maximum allowable or obtainable luminance Y max may differ from.

[0053] A first predetermined value Y min (corresponding to the minimum tolerance of the imaging device concerned) is the null value (0 cd / m 2 ) or a relatively low value (e.g. 100 cd / m 2 ~400cd / m 2 , here 200cd / m 2 ).

[0054] In the first embodiment, the peak luminance providing module 11 provides a peak luminance Y peak (as an input for the mapping unit 10). peak is selected by, for example, a user of the image processing device 1. peak may be input by a user, for example, via a user interface of the aforementioned electronic device, in which case the user interface then converts the input peak luminance Y peakto the mapping unit 10 (via the peak luminance providing module 11). According to another possibility, the peak luminance Y peak may be stored in the storage unit of the image processing device 1, in which case the peak luminance providing module 11 then retrieves the peak luminance Y peak can be read out

[0055] In the second embodiment, the peak luminance providing module 11 provides a peak value Y peak For example, input image I sdr Based on the luminance values ​​associated with each pixel in the input image I sdr as a function of the content of (i.e., the pixel value here).

[0056] An exemplary embodiment of such a peak brightness providing module 11 is described below with reference to FIG.

[0057] The static mapping module 12 calculates (for each pixel p) an input luminance value L sdr the intermediate luminance value L m Convert to.

[0058] In this case, the mid-luminance value L m also depends on the image content (i.e., the input image I sdr (also for pixel values ​​of peak luminance Y peak (This mapping is therefore considered “static”) by applying a function that does not depend on the input luminance values ​​L sdr is obtained from L m =h(L sdr ) However, the function h can be adapted to other conditions, e.g. hdr depends on the type of display it is intended to be displayed on.

[0059] In the currently described example, the static mapping module 12 calculates the intermediate luminance value L m is configured to be obtained as follows:

number

[0060] The central tendency assessment module 14 estimates the intermediate luminance values ​​L generated by the static mapping module 12. m The central tendency value L (for the entire image currently being processed) med This value of central tendency L med is therefore the intermediate luminance value L associated with each pixel of the image being processed. m (i.e., based on the input image I sdr The input luminance value L of all pixels sdr ) generated by the static mapping module 12 based on

[0061] In this example, the value of central tendency L med is the mean luminance value L of all pixels in the image being processed. m This is the median value.

[0062] The central tendency correction module 15 corrects the central tendency value L generated by the central tendency evaluation module 14. med Especially the peak luminance Y peak Based on the correction, the corrected value of central tendency L' med The method is configured to obtain

[0063] Here, to be precise, the central tendency correction module 15 calculates the central tendency value L med Peak luminance Y peak Scale down based on the central tendency correction value L' med The method is configured to obtain

[0064] In addition, in this example, the central tendency correction module 15 also calculates a corrected value of central tendency L' equal to zero, for example by replacing the scaled-down value of central tendency with a predefined (infinitesimal) value L0 if the scaled-down value is less than this predefined value L0.med (to avoid errors in determining the exponent, as described below).

[0065] Therefore, in this example, the central tendency correction module 15 calculates the corrected value of central tendency L' med is specified as follows: L' med =max(L med * Y peak / Y max, L0) where max(I,I') is an operator that generates the maximum value between numbers I and I'.

[0066] The index determination module 16 determines the index γ as a function of the peak luminance value Y peak and the central tendency value of the luminance of the image being processed, here the central tendency corrected value L' generated by the central tendency correction module 15. med The method is configured to identify the target object based on the following:

[0067] In this embodiment, the index determination module 16 determines the central tendency (here corrected) value L' med For the identified exponent γ, the peak value Y peak The saturation voltage is configured to increase as a function of

[0068] In the presently described example, exponent determination module 16 is configured to determine the exponent γ as follows:

number

number

[0069] Therefore, in this example, the exponent is the sum of the first term and the second term δ, where the first term is the input image I sdr The central tendency value of the luminance (here corrected) L'med and / or the second term δ decreases as a function of the peak luminance Y peak where the second term δ is the value of the central tendency correction L' med In this case, the second term δ is a non-positive number and / or the peak luminance Y peak is the maximum brightness Y max When it is equal to , it is equal to zero.

[0070] The adaptive mapping module 18 calculates the output image I hdr For each pixel p' in hdr Let L be the intermediate luminance value (of the corresponding pixel p). m , peak value Y peak , and an exponent γ.

[0071] In this case, the adaptive mapping module 18 more precisely maps the intermediate luminance value L m Here, the mid-luminance value L m is constructed to raise a number proportional to the exponent γ to a power.

[0072] The adaptive mapping module 18 here maps the generated output luminance values ​​L hdr is in a second numerical range [y0, y1], i.e., in particular, the peak luminance Y peak The upper limit y1 is, for example, designed to be lower than or equal to the upper limit y1, which depends on the peak luminance Y peak The upper limit y1 depends on, for example, the peak luminance Y peak In this case, the upper limit y1 is proportional to the peak luminance Y peak and maximum brightness Y max That is, y1=Y peak / Y max It is.

[0073] In the example described here, the adaptive mapping module 18 calculates the output luminance value L hdr is configured to specify the following:

number

[0074] In practice, this means first calculating the intermediate luminance value L m With coefficient m scale (

number

[0075] According to a possible variant, the adaptive mapping module 18 calculates the intermediate value L using the exponent γ as described above. m , but additionally using a specific module to scale (e.g., scale down) the exponentiated value to produce the output luminance value L hdr is in a second numerical range [y0, y1], i.e., in particular, the peak luminance Y peak It may also be made lower than (or equal to) an upper limit y1 which depends on

[0076] FIG. 3 shows curves representing examples of continuously increasing functions used by mapping unit 10.

[0077] Curve C1 (shown by the dashed line) is peak =Y max (That is, here Y peak =1000cd / m 2 ) the continuously increasing function f Ymax Represents.

[0078] Curve C2 (shown as a solid line) is peak =0.4.Y max (That is, here, Y peak =400cd / m 2 ) the continuously increasing function f Ypeak Represents.

[0079] For ease of comparison, curve C' (shown as a dashed line) has the same shape (i.e., curvature) as curve C1, but its upper limit is equal to the upper limit of curve C2. (In other words, curve C' is a function of 0.4.f Ymax Represents.)

[0080] Therefore, as is evident from FIG. 3, here, adjusting the peak luminance not only proportionally scales down the continuously increasing function used by the mapping unit 10, but also scales down the shape of the curve representing this continuously increasing function, i.e., the luminance in a second numerical range [y0, y1] (with the upper limit y1 being the peak luminance Y peak It has been proposed that the way the light spreads across the

[0081] In other words, two peak luminance values ​​Y' peak , Y'' peak For (continuously increasing) function f Y’peak (Peak luminance Y' peak The curve representing the y-axis (the output luminance value L hdr ), i.e. along the x-axis (corresponding to the input luminance value L sdr (corresponding to Y’’peak (Peak luminance Y'' peak (used by Y’peak and f Y’’peak are not proportional to each other.

[0082] Therefore, for a value x different from the maximum value x1 from the first range [x0,x1] (e.g. x=x0+0.5.[x1-x0]), the function f Ypeak The gradient of f Ypeak (x) and the maximum value x1 function f Y’peak The gradient of f Y’peak The ratio with (x1) is Y peak It changes depending on (f Y’peak is a function f Ypeak corresponds to the derivative of the function f Ypeak The gradient of a function f YpeakThe slope of the curve representing

[0083] Two mutually different peak luminance values ​​Y' peak , Y'' peak For peak luminance Y' peak The function f used for Y’peak and peak luminance Y'' peak The function f used for Y’’peak In this case, (function f at x) Y’peak The gradient of the function f at the maximum value x1 Y’peak (ratio between the gradient of f Y’peak (x) / f Y’peak (x1) is a function f Y’’peak The gradient of the function f at the maximum value x1 Y’’peak (ratio between the gradient of f Y’’peak (x) / f Y’’peak It is different from (x1).

[0084] Also, for a value x (e.g., x=x0+0.5.[x1-x0]) from the first range [x0,x1] that is different from the maximum value x1, the relative distance to the linear mapping is the peak luminance Y peak A linear mapping maps an input value x to y0 + (y1 - y0) (x - x0) / (x1 - x0), and therefore the relative distance to a linear mapping is |y0+(y1-y0)(x-x0) / (x1-x0)-f Ypeak (x)| / (y1-y0), where |z| is the absolute value of z (i.e., the modulus).

[0085] In other words, two peak luminance values ​​Y' peak , Y'' peak For function fY' peak and the upper limit y'1 is the peak luminance Y' peak is used for the function fY'' peak and the upper limit y''1 is the peak luminance Y'' peak Note that the peak luminance Y' is used for peak The relative distance to the linear mapping d' when is used is the peak luminance Y'' peakUnlike the relative distance to the linear mapping d'' when d'=|y0+(y'1-y0)(x-x0) / (x1-x0)-f Y’peak (x)| / (y'1-y0) d''=|y0+(y''1-y0)(x-x0) / (x1-x0)=f Y’peak (x)| / (y''1-y0) It is.

[0086] In this case, x0=y0=0,y'1=Y' peak / Y max And y''1=Y'' peak / Y max In this case, for any value x (e.g., x=x0+0.5.[x1-x0]) from the first numerical range [x0,x1] that is different from the maximum value x1, [x / x1=Y max. f Y’peak (x) / Y' peak ] is [x / x1-Y max. f Y‘’peak (x) / Y'' peak ] (This is different from the function f Y’peak and f Y’’peak are not proportional to each other.)

[0087] The relative distance to the linear mapping (defined above for a number x taken from a first range [x0,x1] and different from the maximum value x1, e.g. x=x0+0.5.[x1-x0]) is e.g. the peak luminance Y peak may increase as a function of

[0088] For example, the curvature of the applied continuously increasing function may be made to increase as a function of the peak value.

[0089] The shape of the curve is the peak value Y peak When a wide dynamic range is available (i.e., peak luminance Y peak, which significantly extends the highlights when the peak luminance Y peak This helps to avoid excessive crashing of midtones and shadows to black when is low.

[0090] In the embodiment of the mapping unit 10 described above, this is the peak value Y peak However, in other embodiments, the peak luminance Y peak The change in the shape of the curve representing a continuously increasing function according to the peak luminance Y peak By varying other parameters of the continuous increasing function according to peak This can be obtained by modulating the effect of the static mapping module 12 depending on the

[0091] Peak Luminance Y peak This modulation of the effect of the static mapping module 12 in response to the peak luminance Y peak Instead of the change in the exponent γ as a function of peak In addition to varying the exponent γ as a function of

[0092] In a possible embodiment, the exponent γ is peak and / or the intermediate luminance value L m is the input luminance value L sdr From Y peak A function λ with parameters Ypeak is obtained by applying L m = λ Ypeak (L sdr )

[0093] In addition, the function λ Ypeak For example, any L in the first numerical range [x0, x1] sdr Regarding λ Ymin (L sdr )=L sdr(i.e., the function λ Ymin is the identity function).

[0094] Function λ Ypeak For example, n-th degree polynomial, λ Ypeak (x)=m0.x n +m1.x n-1 +...+m n-1 .x+m n where parameters m0, m1, ..., m n is λ Ymin (L sdr )=L sdr In this goal, the parameter m i At least some of peak depends on m for i other than 1 i (Y min )=0, and m i (Y min )=1.

[0095] However, other types of functions can be used, such as spline-based functions.

[0096] Peak Luminance Y peak Y min This, together with the fact that the exponent γ decreases (e.g., toward 1) when peak =Y min Linear scaling with respect to (i.e., L hdr =(Y peak / Y max ).L sdr ) can be converged to

[0097] FIG. 4 shows a possible embodiment of the peak luminance providing module 11 used in the mapping unit 10.

[0098] In this embodiment, the peak luminance providing module 11 includes a brightness estimating module 110 and a peak luminance identifying module 112 .

[0099] The brightness estimation module 110 estimates the brightness of an input image I sdr Therefore, here we use the input image L sdr The input luminance value of the pixel in L sdr The method is configured to evaluate a brightness value μ derived from

[0100] To be precise, in this example, the brightness value μ is calculated by the static mapping module 12 (as described above, the input luminance value L sdr The intermediate luminance value L m It is calculated from:

[0101] According to a first possibility, the brightness value μ is a function of the luminance value (i.e. here the intermediate luminance value L m ) (e.g., a predetermined percentile equal to or greater than the 60th percentile). In this example, the brightness value μ (here, the mid-luminance value L m ) the 80th percentile of the luminance values.

[0102] In other words, in this case, the brightness value μ is the value at which a certain percentage of the pixels of the image (here 80%) have a brightness below that value (here the intermediate brightness L m ) are (here intermediate) luminance values.

[0103] According to the second possibility, the brightness value μ is (here the mid-luminance value L m Weighted median of luminance values ​​m w This weighted median m w is calculated, for example as described herein, giving greater weight to bright pixels.

[0104] It is proposed here that one can define N bins spanning all possible luminance values ​​(here all possible intermediate luminance values), with a bin i ∈ {1,...,N} corresponding to pixels with luminance values ​​between V(i) and V(i+1), and H(i) being the number of these pixels with luminance values ​​between V(i) and V(i+1) (hence V(1) corresponds to the smallest possible luminance value, V(N+1) corresponds to the largest possible luminance value, and V(j) for any j ∈ {1,...,N}). <V(j+1)である)。

[0105] A weight W(i) is assigned to each bin i as follows:

number

[0106] The brightness estimation module 110 therefore calculates the cumulative weighted histogram C as follows: For every i ∈ {1,...,N},

number

number

[0107] The brightness estimation module 110 therefore calculates the weighted median m w where k is the first (i.e., lowest) element in {1,...,N} such that C'(k) >= 0.5.

[0108] The peak luminance determination module 112 determines the peak luminance Y peakbased on the brightness values ​​μ generated by the brightness estimation module 110.

[0109] Here, the peak luminance Y peak It is proposed that σ decreases as a function of the brightness value μ. This allows to improve viewing comfort and / or improve power consumption. Brighter images lead to higher brightness values ​​and therefore lower peak luminance Y peak , which ensures that the average brightness of the resulting image is not uncomfortably high.

[0110] In the example described here, the peak luminance identification module 112 determines the peak luminance Y peak is configured to calculate If μ ≥ τ,

number

[0111] FIG. 5 shows an example of an image processing device 21 according to another embodiment of the present invention.

[0112] As will be apparent from the following description, the image processing device 21 defines a first dynamic range Δ hdr An input image I′ having (e.g., high dynamic range, or HDR) hdr The second dynamic range Δ sdr (e.g., standard dynamic range, or SDR) sdr The second dynamic range Δ sdr is the first dynamic range Δ hdr (strictly) lower.

[0113] First Dynamic Range Δ hdr Input image I′ having hdr The first dynamic range Δ hdr Lower second dynamic range Δ sdr Output image I′ having sdr Such a process of converting to is commonly called "tone compression."

[0114] The image processing device 21 can for example be used to process images generated by the image processing device 1 described above with reference to FIG. 1, in particular to process the original input image I sdr (a process commonly known as a "round trip").

[0115] The image processing device 21 may in fact be implemented by electronic equipment including a processor and a memory for storing program code instructions adapted to perform the operations and functions of the modules and units described below when the relevant program code instructions are executed by the processor.

[0116] Input image I' hdr is the input image I' hdr For each pixel p, we use a set of pixels (generally a matrix of pixels) to calculate multiple component values ​​r p , g p , b p It is represented by:

[0117] In this example, the input image I' hdr are the three color components r of each pixel p p , g p , b p (i.e., the red component r p , green component g p , and the blue component b p However, the input image I' hdr Other representations may also be used for, for example, the luminance component y for each pixel p p and two chromaticity components u p , v p is used.

[0118] As can be seen in FIG. 5, the image processing device 21 processes an input image I' hdr , i.e., the input image I' hdr The component value r represents p , g p , b p In practice, the input image I' hdr The component value r represents p , g p , b p may be received from other modules of the electronics or from external electronics (through a communication circuit cooperating with the input module 22).

[0119] The image processing device 21 also receives the input luminance value L' hdr the component value r (of the relevant pixel p) p , g p , b p The image generation module 24 is configured to generate the image based on the

[0120] The image creation module 24 receives the input luminance value L' hdr The component value r p , g p , b p In this case, several steps are required to generate the - Inverse Optical Electrical Transfer Function (i.e. OETF -1 ) application, - Calculating luminance from RGB representation can be executed.

[0121] The input luminance value L' generated by the image creation module 24 by the step of applying the inverse photoelectric transfer function hdr is the input image I' hdr Represents the linear luminance component of

[0122] The image processing device 21 receives an input image I' hdr The input luminance value L' associated with any pixel p of hdr The output image I' sdrThe output luminance value L' associated with the corresponding pixel p' in sdr The mapping unit 30 is designed to convert the

[0123] As described below, the mapping unit 30 calculates a continuously increasing function φ that maps a first range of values ​​[y0, y1] to a second range of values ​​[x0, x1]. Ypeak whereby the upper limit y1 of the first numerical range is determined by a continuously increasing function φ Ypeak The upper limit y1 may be selected by a user of the image processing device 21 or may be determined by, for example, the input image I' hdr Peak luminance Y can also be received as metadata associated with peak It is adjustable depending on (and relative to)

[0124] Function φ Ypeak is an increasing function if, for any pair (z1, z2) in the first numerical range [y0, y1], z1≦z2, φ Ypeak (z1)≦φ Ypeak This means that (z2).

[0125] As will be explained further below, the continuously increasing function φ used Ypeak is the adjustable peak value Y peak and therefore varies with φ Ypeak It is written as follows.

[0126] In this case, the continuously increasing function φ Ypeak is each input luminance value L' hdr (included in the first numerical range [y0,y1] and related to a particular pixel p) is the output luminance value L' sdr (The second numerical range [x0, x1] contains the output image I' sdr , which relates to its corresponding pixel p' in

[0127] A possible embodiment of the mapping unit 30 is described below with reference to FIG.

[0128] The image processing device 21 outputs an output image I' sdr For each pixel p' in p’ , g' p’ , b' p’ Let p′ be the output luminance value L′ generated by the mapping unit 30 for the associated pixel p′. sdr and the input image I' hdr The component value r for the corresponding pixel p in p , g p , b p The image generation module 26 is configured to identify the object based on the

[0129] Component value r' p’ , g' p’ , b' p’ Output luminance value L' sdr and the component value r p , g p , b p To identify the object based on, for example, the image generation module 26 may - Output luminance value L' sdr Using the component value r p , g p , b p , whereby the scaled component values ​​are converted to output luminance values ​​L' sdr ), - performing color correction (e.g. by adjusting saturation), - a possible step of converting from a representation in the ITU-R BT.2020 standard to a representation in the ITU-R BT.709 standard, - possible steps of applying the optical-electrical transfer function (i.e. OETF) Execute.

[0130] In this example, the component value r' p’ , g' p’ , b' p’ is the output image I' sdr For each pixel p', the three color components r' p’ , g' p’ , b' p’(i.e., the red component r' p , green component g' p’ , and the blue component b' p’ ) However, the output image I' sdr Other representations for p′ may be used, e.g., the luminance component y′ for each pixel p′ p’ (possibly, in this case the output luminance value L' generated by the mapping unit 30 for the pixel p' concerned) sdr ) and the two chromaticity components u' p’ , v' p’ is used.

[0131] The image processing device 21 calculates the component value r' p’ , g' p’ , b' p’ , for example by a display module of the aforementioned electronic device, on a screen of the electronic device. sdr is used to display the image or, in a modified form, is transmitted to an external electronic device (using the electronic device's communication circuitry).

[0132] FIG. 6 shows a possible embodiment of the mapping unit 30.

[0133] The mapping unit 30 includes a peak luminance providing module 31, an index identifying module 32, an adaptive mapping module 34, and a static mapping module 36.

[0134] The peak luminance providing module 31 provides a peak luminance Y peak (as an input for the mapping unit 30). peak is selected, for example, by a user of the image processing device 21. peak can be input by a user, for example, via a user interface of the electronic device described above, in which case the user interface will then peak (via the peak luminance providing module 31) to the mapping unit 30. According to another possibility, the peak luminance Ypeak can be stored in the storage unit of the image processing device 21, in which case the peak luminance providing module 31 then provides the peak luminance Y peak can be read out from the storage unit. According to another possibility, the peak luminance Y peak is the input image I' hdr , and the peak luminance providing module 31 then derives the peak luminance Y by parsing these metadata. peak can be obtained.

[0135] When trying to obtain a round trip of an image previously processed by the image processing device 1 described with reference to FIG. 1, the same peak luminance Y peak should be used.

[0136] The exponent determination module 32 determines the exponent γ as the input luminance value L' hdr (Current input image I' hdr ) generated by the image creation module 24) and the peak luminance Y peak The method is configured to identify the target object based on the following:

[0137] In this example, the exponent determination module 32 assigns an exponent to the input luminance value L' hdr The central tendency value (here, the median) of the peak luminance Y peak , possibly corrected to be at least equal to the minimum value L0.

[0138] Therefore, in the example described herein, the index identification module 32 is configured to calculate a scaled down, corrected value of central tendency as follows:

number

[0139] The index identification module 32 further determines the index γ as a function of the (here scaled down and corrected) value of central tendency and the peak intensity Y peak The method is configured to determine the eigenvalue as a function of

[0140] In the example described herein, the index identification module 32 is configured to calculate the index as follows:

number

number

[0141] Therefore, in this example, the exponent is the sum of the first term and the second term δ, where the first term is the input image I' hdr and / or the second term δ decreases as a function of the (here scaled down and corrected) value of the central tendency of the luminance of the peak luminance Y peak where the first term is the peak luminance Y peak One may notice that the second term δ does not vary as a function of the value of the central tendency. In this case, the second term δ is a non-positive number and / or the peak luminance Y peak is the maximum brightness Y max When it is equal to , it is equal to zero.

[0142] The adaptive mapping module 34 calculates the intermediate luminance value L m Let L' be the input luminance value. hdr and the exponent γ (generated by the exponent determination module 32).

[0143] To be precise, the adaptive mapping module 34 now converts the input luminance values ​​L' hdr by raising to the inverse of the exponent γ (generated by the exponent determination module 32)m The method is adapted to obtain

number

[0144] The static mapping module 36 (for each pixel of the image being processed) (output image I' sdr (for its corresponding pixel in sdr the intermediate luminance value L m (generated by the adaptive mapping module 34 as described above) and the peak luminance Y peak In addition to being based on, here, the determination is configured to be based on the exponent γ.

[0145] In the illustrated example, the static mapping module 36 calculates the output luminance value L' sdr is configured to calculate the following:

number

[0146] Therefore, the input luminance value L' hdr is equal to the upper limit y1 (and here the adjustable peak luminance Y peak and maximum brightness Y max (corresponding to the ratio of Y peak is used, the mapping unit 30 outputs an output luminance value L' equal to 1 (i.e., equal to the highest value x1 in the second numerical range [x0, x1]). sdr Generate. [Aspect 1] An image processing apparatus (1) for converting an input image having a first dynamic range into an output image having a second dynamic range higher than the first dynamic range, the apparatus comprising: sdr ) to the output luminance value (L hdr a mapping unit (10) for converting The mapping unit (10) is configured to apply a continuously increasing function to map a first range of values ​​to a second range of values, the highest value (x 1 ) is the upper limit (y 1 In an image processing device (1) that is mapped to The mapping unit (10) is a peak luminance (Ypeak ) according to the upper limit (y 1 ). [Aspect 2] The mapping unit (10) calculates the peak luminance (Y peak 2. The image processing device of claim 1, further comprising a module (11) for determining [Aspect 3] The peak luminance (Y peak 3. The image processing apparatus of claim 2, wherein the module for identifying a peak luminance is configured to identify the peak luminance based on luminance values ​​respectively associated with pixels of the input image. [Aspect 4] The peak value identification module (11) evaluates the brightness value (μ) of the input image to determine the peak luminance (Y peak 4. The image processing device according to aspect 2 or 3, configured to determine [Aspect 5] The mapping unit (10) calculates the peak luminance (Y peak 5. The image processing device according to any one of aspects 1 to 4, wherein the image processing device is configured to change at least one parameter (γ) of the continuously increasing function in response to [Aspect 6] The mapping unit (10) maps the exponent (γ) to the peak luminance (Y peak 6. The image processing device according to any one of aspects 1 to 5, further comprising: a module (16) for specifying based on the exponent (γ), and a module (18) for raising the exponent (γ) to a power. [Aspect 7] The module (16) for determining the exponent (γ) determines whether the exponent (γ) is greater than the peak luminance (Y peak 7. The image processing device of embodiment 6, wherein the image processing device is configured to increase the image density as a function of [Aspect 8] The mapping module (10) converts the input luminance values ​​(L sdr ) to the midpoint luminance value (L m ) to the peak luminance (Y peak 8. The image processing device according to any one of aspects 1 to 7, comprising a module (12) configured to transform by applying a mapping function that depends on [Aspect 9] A method for converting an input image having a first dynamic range to an output image having a second dynamic range higher than the first dynamic range, comprising: sdr ) to the output luminance value (L hdr ), the step of converting includes the substep of applying a continuously increasing function that maps a first range of values ​​to a second range of values, thereby obtaining a maximum value (x 1 ) is the upper limit (y 1 ), The upper limit (y 1 ) is the peak luminance (Y peak ) adjustable depending on the [Aspect 10] The peak luminance (Y peak 10. The method of claim 9, comprising determining a pixel value of the input image as a function of the pixel value of the input image. [Aspect 11] The peak luminance (Y peak The step of identifying the peak luminance (Y peak 11. The method of embodiment 10, comprising the substep of determining (μ) as a function of said value (μ). [Aspect 12] The parameter (γ) of the continuous increasing function is the peak luminance (Y peak 12. The method according to any one of aspects 9 to 11, wherein the above can be varied depending on the above. [Aspect 13] 1. An image processing apparatus (21) for converting an input image having a first dynamic range to an output image having a second dynamic range lower than the first dynamic range, comprising: hdr ) to the output luminance value (L' sdr ), comprising a mapping unit (30) for converting The mapping unit (30) is configured to apply a continuously increasing function that maps a first range of values ​​to a second range of values, the upper limit (y 1 ) is the highest value (x 1 In an image processing device (21) that is mapped to The mapping unit (30) determines the upper limit (y 1 ) is the peak brightness (Y peak ) ...).

Claims

1. An image processing apparatus (1) for converting an input image having a first dynamic range into an output image having a second dynamic range higher than the first dynamic range, the apparatus comprising: ｓｄｒ ) to the output luminance value (L ｈｄｒ ), The mapping unit (10) is configured to apply a continuously increasing function to map a first range of values ​​to a second range of values, the highest value (x １ ) is the upper limit (y １ In an image processing device (1) in which The mapping unit (10) comprises a module (11) for identifying a peak luminance (Y peak ) as a function of pixel values ​​of the input image, the module (11) being configured to evaluate a brightness value (μ) of the input image to identify the peak luminance (Y peak ) as a function of the value (μ), and the mapping unit (10) is configured to identify the peak luminance (Y peak ) as a function of the pixel value (μ). ｐｅａｋ ) depending on the upper limit (y １ ).

2. The mapping unit (10) calculates the peak luminance (Y ｐｅａｋ 2. The image processing device according to claim 1, configured to vary at least one parameter (γ) of the continuously increasing function in dependence on

3. The mapping unit (10) maps the exponent (γ) to the peak luminance (Y ｐｅａｋ 3. The image processing device according to claim 1, further comprising a module (16) for determining based on the exponent (γ) and a module (18) for exponentiating the exponent (γ).

4. The module (16) for determining the index (γ) is configured to determine whether the index (γ) is greater than the peak luminance (Y ｐｅａｋ 4. The image processing apparatus of claim 3, wherein the image processing apparatus is configured to increase as a function of

5. The mapping unit (10) converts the input luminance values ​​(L ｓｄｒ ) to the intermediate luminance value (L ｍ ) to the peak luminance (Y ｐｅａｋ 5. An image processing device according to claim 1, further comprising a module (12) adapted to transform by applying a mapping function depending on:

6. The image processing device according to claim 1, wherein the peak luminance (Y peak ) decreases as a function of the brightness value (μ).

7. A method for converting an input image having a first dynamic range to an output image having a second dynamic range higher than the first dynamic range, comprising: ｓｄｒ ) to the output luminance value (L ｈｄｒ ), the step of converting includes the substep of applying a continuously increasing function that maps a first range of values ​​to a second range of values, thereby obtaining a maximum value (x １ ) is the upper limit (y １ ) in which The method includes the steps of: determining a peak luminance (Y peak ) as a function of pixel values ​​of the input image, the determining step including the sub-steps of evaluating a brightness value (μ) of the input image; determining the peak luminance (Y peak ) as a function of the value (μ); and determining the upper limit (y １ ) is the peak luminance (Y ｐｅａｋ ) adjustable depending on the

8. The parameter (γ) of the continuous increasing function is the peak luminance (Y ｐｅａｋ 8. The method of claim 7, wherein the temperature is variable depending on the temperature.

9. The method of claim 7 or 8, wherein the peak luminance (Y peak ) decreases as a function of the brightness value (μ).

Images