TR201906704T4 - Methods and devices for creating code mapping functions for encoding an hdr image, and methods and devices for using such encoded images. - Google Patents

Abstract

In order to allow better encoding of already starting to appear as high dynamic range images for use in full high dynamic range technical systems (including an HDR display screen and, for example, an HDR film in an HDR grading application), we include the following steps of encoding a high dynamic range image. We invented the method of: inputting pixel colors of an input high dynamic range image, wherein pixel colors have information of a luminance and a chromaticity; applying an inverse of a mapping function to derive a luma code (v) of the luminance of a pixel color, in which p is a tuning constant and v is the luma code corresponding to a luminance to be encoded, defined as (I). the partial function and predetermined including a second partial mapping defined as L = LmPY in which Lm is a peak luminance of a predefined reference display screen and the gamma is a constant preferably equal to 2.4; Outputting a matrix of pixels having a color coding containing luma codes.

Description

DESCRIPTION CODE MAPPING TO CODE AN HDR IMAGE METHODS AND DEVICES TO CREATE THEIR FUNCTIONS AND TO USE SUCH CODED IMAGES METHODS AND DEVICES TECHNICAL FIELD Invention one (for example, a photograph), but preferably more (video) High Dynamic Range images), either naturally we just have an HDR brightness in the sense that we need an encoding for the appearance image (typically with high peak brightness such as 5000 nits and so on, for example, display screen with important objects across multiple brightness levels to deep black an optimal image for display on) or by a binary encoding is relevant: and here, in addition to the HDR image appearance, we have a corresponding We encode the LDR image appearance. In addition, coding preferentially, into the current image or e.g. Blu-ray disc storage or HDMI cable connections or other video transmission or storage systems, etc. to fit within the video coding frameworks of existing technology. HDR Video (or even still image) encoding has been a daunting task until now. has been the task, and the typical belief is that the person is either encodings that directly encode scene brightness) LDR range Significantly more bits are required to encode the brightness above. or the person needs some two-layered approach is hearing, and here, for example, in addition to the reflection image of an object, a There are illumination enhancement images or similar degradation strategies. Philips recently proposed a much simpler single-image approach. exists and this is a completely new direction and it can only be imagined. Not that it was difficult before, but at the same time, it's actually hard to do it. leads to the solution of many technical issues, but in practical matters, and specific In this context, the current patent application text does not use such a coding technology. 28834.1284 and different reproduction scenarios (one HDR display and one LDR for the image that appears at least as a reality/optimal for display) some of the building the whole framework around it, like artistic grading. teaches its parts. image(s) have a higher brightness contrast compared to older LDR coding We describe it as having a contrast ratio of 10,000:1 or more (for example, a contrast ratio of 10,000:1 or more). Manipulating the image until achievable visualization through encoding rates All components of the chain can be achieved; and captured object brightness is 1000 nits or more specifically, typically above 1000 nits They can be reproduced and this is the result of a lit lamp or sunny outside. given to the reproduction environment that creates some desired appearance of the space) and/Or the visualization of such image(s) is HDR (e.g., images they have to be affordable and here, they are high quality HDR for visualization and, preferably, in a technically easy-to-use We would like to mention that they contain sufficient information within the approximation, and that the image(s) are the most are visualized on display screens with a peak brightness of at least 2000 nits or are intended to visualize them (however, this is typically after appropriate color mapping, for example at 100 nits peak brightness LDR does not indicate that they cannot be visualized on display screens) BACKGROUND ART Recently, for example, Dolbys's two-layer method Map the HDR image into a lower bit depth image and explains that this results in lower quantization error. F. Banterle et al., 28834.1284 The published article A framework for inverse tone mapping is an old view a reverse tone for better visualization of HDR content on the screen. Explains the mapping operator.

However, as the industry still exists, the amount of data, but also, computational complexity (price of lCs), 'ease' of my answer, whatever they want all important factors such as the versatility of the artist to create A pragmatic HDR video (image) that fits with the requirements (a balance of them) looking for coding. In particular, a dual-layer approach is complex is seen. One would ideally like to use a device that adapts to the old coding, for example DCT-based Being competent as an encoding designer, such as MPEG HEVC encoding will enjoy it. One problem is that this is somewhat contrary to conventional wisdom (however, One can encode a DR image and it will, for example, have a peak of around 100 nits. Specific LDR to be viewed on display screens with brightness and in low light conditions engaging in a technology optimized to contain images An LDR that typically has a large amount of brightness must be something other than the image as a description). This is old LDR image processing/coding systems provide a total contrast of approximately 68:1” (170:1, respectively!) ratio is given for most visible objects (85% for white and 85% for black in reflection) at studio lighting ratio (or, for example, 10:l) which can vary between 5% For example, normally well-illuminated antitypical LDR imaging with a 4:1 It was defined and optimized to work with different scenarios. If one is a peak If one looks at the relative visualization of the luminosities starting in white, the local With the image contrast ratio of a typically older LCD monitor without dimming would have something like 100 nits white and lnit black to match, and typically, day at the same time to have something like a 40:1 proficiency. will have been tracked above the average systems that can be tracked during was considered. In these systems, a standard brightness code allocation of 2.2” gamma for most scenarios of higher scene contrasts It seems to be satisfactory. Although referred to as acceptable errors High brightness coded as bad, although some were made at the time 28834.1284 Visualization of scene areas (e.g. hard clipping) can cause such errors are also acceptable because LDR display screens do not physically They cannot visualize.

However, on an HDR display screen we can make them both bright and colorful. While you want to show off, the person inside is simultaneously exposed to sunny outdoor areas. visible, which may have an illumination ratio of 100:1° or even more, e.g. There is a desire to develop visualization like an indoor scene There are scenarios. In LDR7, these regions are (softly) clipped. as they are to be displayed (typically for those pixels in the image that has already been encoded) maximum code around 255 is difficult to distinguish), an HDR display On the screen, we'll want to make them look both bright and colorful. This such scenes (as if you were actually on holiday in Italy), but even some higher brightness content of some scenes may cause some reflective The creation of reflections still shows a major quality improvement, It will give more natural and extraordinary visualization. If for example a 5000 or 10000 nits display clipping on the screen or niccmlcmc errors etc. If artifacts can be seen as annoying, at least we can see that the image is accurate. We would like to be competent to drive such display screens with its type and so the visualization will be as beautiful as the display screen allows.

However, in order to encode the classical intelligent additional and extreme brightness ranges, one can use (much) more It will need more bits. This is either native to single larger code words (16 bits with one sign bit, 5-bit exponential and 10-bit mantissa) bit Open EXR or all possible object brightness with high precision Wardss, who makes a mathematically charged effort to capture the world LogLuv encoding) or standard LDR range codes (such as HDR a first layer with a classical JPEG approximation of the image) and so on a second layer to enhance pixel brightness to higher brightness (for example, if higher brightness is needed, increase each pixel by 28834.1284 an upscaled image, for example, two images equivalent to a single linear 16-bit code This can occur through multiplication of such 8-bit images. One main practical problem to be solved is that a practical HDR coding technology is designed. time, its a very large range of different HDR images to process In addition to the fact that hardware manufacturers must be competent to They want lower amounts of hits per code word (channel) and each how much, the technology we offer below is also larger bit works with words, we need at least one brightness (or more precisely, a came up with a solution that works well under a 10-bit limit for the luma channel. We leave. In addition, we provide several visualizations within one functional approach. Both color pixel coding and color appearance for the scenario We have developed a framework that can be transformed into a dual-philosophy and here, Instead of this being for every image, at least a second image, it's just the functions It means that co-coding is needed. Researching this path and Through discovery, we see what at first glance would not be a detail and this patent of the appropriate function(s) on the luma axis as described in the reference selection and even the other two in a luminance-independent chromaticity plane make the system actually work with good quality through the coding of the component we discovered that we will bring and here, from now on, the development color plane Free selection (e.g. for wide gamut) is easy within the codec space itself. It offers additional advantages of this encoding such as calculations.

BRIEF DESCRIPTION OF THE INVENTION We need an improved coding of HDR images and, in particular, Starting with this philosophy, especially in today's time, there are still many old LDR system exists there in the field, we achieve some level of comparability. we need it. On the one hand, this is like (I)DCT [level one compliance] Use of existing (decoder)encoder that implements the functionality We want to preserve, but at the same time, we need LDR images. 28834.1284 This means there is a need for second level compatibility with display screens that comes because they can only visualize the LDR (i.e. the correct LDR look under such a display dynamic range capability, e.g. with the dark regions still being sufficiently visible) because it is an addition to older TVs already shipped As such, in the future, a consumer will also have low brightness laptop that `will want to see some rendering of an HDR movie` in its competency or small portables like pad-computers or even ccp tclcfon1ari devices that can have a peak brightness of, for example, 10000 nits in the future. display screens in the range up to advanced HDR display screens a spectrum and all display screens between or around them It will happen. After that, although the display screen may still be old and simple also supply HDR content, for example, via an HDMI or other connection new high-complexity decoding on the set-top box or computer that The color mapping can be serviced by IC and set-top box invented by our It offers any combination of the options we have discussed and described. We In this, in the ideal scenario, the person would receive the same movie or pictures from the content provider. In case of an approximation that will need (cn less) two dcccclcndirmcyc We have converted it simply, (e.g. 100 nits peak In LDR display screen scenarios, such as display screen with brightness one LDR image (to be used) and one HDR image (for brighter displays) We will call it as, but the following applications can also be used if the person is not only an HDR image (for example, on a blu-ray disc and after that, either just They serve a specified range of HDR display screens or A mapping for out-of-range display screens in a standalone approach derives) are useful, and we present the doctrines, and thus, they They fit any prescribed strategy.

So, for a few practical example scenarios, an input is essentially an HDR grade image. to new HDR encoding, for example a new Blu-ray disc or Internet streaming the television IC or applicable video/image coding technologies receiving the video. any image source that is substantially compatible with a variant of 28834.1284 For any connected receiver, we have as a starting point (each What color processing software is it according to any taste of the creator optionally graded and, for example, an OpenEXR-like initial encoded in color coding) and then the current Video and image practically useful for technologies (e.g., such coding Only slightly modified in the normal way to use their technology has been done, but not all buses need to be converted to 12 bits, e.g. Our methods should work with 12-bit hardware, but at the same time, or if only 10 bits per component can be provided, 8-bit systems (even if one accepts some lower quality) on the need for encoding We see that an HDR image and some “LDR” image (its LDR visualization a rating to be used directly for or just some “pseudo- image”, but after additional color processing, via an HDR or It will not be seen to encode an image to be visualized in HDR technology) A system on color components that encodes the brightness of pixel colors. realize its connectivity to the HDR image through functional conversion We have reached the goal (either a single function or a useful and pre-agreed, typically one of a limited set of pre-agreed remaining, a content creation side, for example, of a movie for a television that receives a movie a content, such as a pair of tenses at the beginning and possibly during playback When the field is functionally linked to the side, it eventually occurs should come: for example, if they are variable, they are not constants and many full movies When agreed to encode, these functions operate in a linked memory may be transmitted via a network communication path stored on the device).

Our invention can, for example, be realized at least in the following ways: One method of encoding a high dynamic range image is as follows: 28834.1284 - input pixel colors of an input high dynamic range image and where, pixel colors have a brightness and a chromaticity information; - to derive a luma code (v) of the brightness of a pixel color implementation of an inverse of the mapping function, mapping function P : (iL-41), containing a first partial function defined as where rho (p) is a conformation constant and V It is the luma code corresponding to a brightness (L) to be encoded, its conversion to a gain is L : a second partial function defined as LmPy and where, Lm a predefined reference is a peak brightness of the display screen and gamma (y) is a constant preferably equal to 2.4, - a set of pixels that have a color coding that includes luma codes (v) Outputting the matrix.

After that, this function uses N-bit luma's Y, for example codes [0-1023] with colors having normalized versions, symbol V in the given range (there may be many, for example an Egyptian market place with one person in the shade, but the same at the same time, by the strong sunlight shining through the holes in the roof a second one that is illuminated, but the video can also be used, for example, as a weather map or lasers or artificially stimulated lasers reporting in a disco to ensure that they are encoded with sufficient precision. It can be used as follows. Thus, this function can be intelligently predetermined or it can be determined at work so that, before uniform quantization, that re-allocates brightnesses to lumas. This brightness mapping function (or more accurately, a reference display screen, in the form of a starting point) which maps luma codes at [0,1] to visualizable luminances on When the corresponding Electro-optical transfer function (EOTF) is designed, we We have a few technical behavioral criteria in mind. First, although 28834.1284 (partial) logarithmic behavior to encode many tens of brightness While proficiency is a nice trait to have, it certainly takes away the brilliance. some linear interval does not just have a logarithmic representation (strictly speaking, it just There is only one desired factor that takes into account image processing reasons. without processing to obtain the image appearance and at the same time, For optimal encoding there will be only one truncated adaptive log-function in a non-existent way). On the contrary, this function emerges from long experience. It was designed after several considerations and in terms of these considerations, In practice, it adequately captures all the necessary appearances occurring in HDR scenes. To realistically represent a defined universal linear reference Once you are competent to define the brightness range (for example, 0.0001-5000 nits) then this is an optimal for all criteria (at least good HDR quantization precision) It was seen as a behavioral function. Officially our reference EOTFs Note that they also have exponential parts, but one also In this way, one can see and describe this as “logarithmic” behavior. Now, that having a reference range, some for bright luminances (or lumas) Being would be the most useful state. Indeed, mathematically as l+x approximation exp(x), as we call it in the dark, we predominantly use gamma behavior and this gives us an easily controllable function. means we have, because we need in the dark with the parameter rho We can adjust the sensitivity (gamma to 1 at maximum) The given additional format becomes controlling because the maximum code that returns the representation is corresponds to the maximum displayable brightness). This is HDR image only It wouldn't be useful to have a single master EOTF for encoding, but the same At the same time, if we have different classes of HDR images, gamma it becomes useful. In fact, if we have a single optimal solution that is reasonable for all scenarios If we do not decide to implement the function, we will use the EOTF° (and the corresponding inverse We can set OETF7). For example, the expert reader is in a corner of a dark basement. On the one hand, the image appears darker to the viewer's eye, especially in dark environments. that it is adapted to see gray values and, on the other hand, is potentially more relevant 28834.1284 that it has darker gray values (poorly reflected by various reflections). since all these illuminated objects fall into darkness) and especially, If there are only lamps, bright ones may not need ultimate sensitivity. can imagine. Conversely, if the person often sees very bright luminances (e.g., 1000 and 3000 nits) looking at a sunny scene, ultimate sensitivity is needed for darkness may not be heard, and the visual system typically ignores this to a large extent. will. This can all be done easily with our gamma and rho format parameters. adjustable and they control pattern behavior in the dark versus bright and also at the bottom of each luma step, especially the luma code region. The amount of only noticeable differences corresponding to the region can be controlled (also called encoding resolution or precision). One fixed pre-agreed parametric function, 0 newly found what the codes really mean, but adaptability is more generic In case of pre-agreed function to decode Communicating its parameters to the recipient since he/she will use it does not need, for example, a human evaluator or an automated imager required as selected at the receiver end by the analysis program. parameters (at least one of the peak luminance Lm, rho and gamma) e.g. co-coding on a memory such as a BD disc, meta-coding in an image signal co-transmission as data at playback time via the path of another transmission any of a number of possible mechanisms, such as re-correction The buyer will be contacted by.

But at the same time, another interesting thing happens with this technology: HDR When an encoding of the input is created, this function also lower dynamic range image (which one of a small number of bits can be coded better, but a more accurate technical formulation There will be at least some more significant sub-range or sub-range of the image histogram closer to medium gray compared to the original inputted main HDR image will be encoded with luminances as ) but on the other hand, the encoded LDR The image still has all the information to recover the original master view 0 28834.1284 At least one higher degree of approximation is preferably is preserved and thus contains all the important data of that HDR image.

Thus, if the EOTF parameters are chosen well, the resulting encoded image An LDR display screen that gives a reasonably good quality LDR appearance without additional optimizing color processing to visualize on or from some additional second-order optimization transformations (e.g., LDR representation mapped to a different set of priorities) can be used directly.

When a single main EOTF function is coded (if we If we use a single rather than one selected from a sequence), visualization That is, the function 0 can be used generically, without taking into account its specifics. we can define, for example, however, this is at least the image of the HDR view It is still mostly a system that has adequate coding of its textures only in the mind. through a display screen referred to as the reference display screen is defined. But preferably, at the same time, we even have an LDR view we encode, preferably along the luma axis in the validation sub-ranges We still position the various captured scene objects and thus, directly When displayed directly, the LDR view is chosen by the content creator. It will appear as done. But we also have a a given track with additional optimized luma allocation function We already code for the application, luma codes such as 15 lux a reference brightness or surrounding medium gray (18%), such as 10 nits reflectance) characterized by a reference environmental brightness, such as a They give the best quality for viewing in dark or low-light conditions. This is some Incorporating extra gamma function in applications, such as a 1.25 It can be done via . Ultimately, this EOTF in our core applications, our entire desired part as an optimal final gamma to change gaina It corresponds to the time when we model gammas. All variations can be provided achieve the best image object quality for a given amount of bits or by other means, e.g. just for relatively few bits, like 12 bits While achieving good quality, we turned out to be our preferred choice, 28834.1284 For example, a large color plane in the keromatic direction similar to UCS If we use it, we use wide color gamut capacities at the same time.

We also wonder how this inverted EOTF goes from dirt to luma codes. mapping (as well as a camera or color processing luminances from the software to the resulting electrical codes, which are typically digital 2^N times which maps the optical measurements and where N is the amount of bits can be viewed as a unit range [0,1]) as one of its corresponding EOTF can be viewed as the corresponding optimized (theoretical) HDR reference representation how it works best with the screen (the theoretical code directly have the same characteristics because they are capable of visualizing We have discovered it with a real demonstration screen. Independently, on the final display screen on which the video needs to be visualized, A unique non-jamming approach to video with its luma codes has to be defined in (in principle, the code creation side then You can use any principle to allocate codes, but we stick to simplicity. In terms of additional evaluation, it is only seen from the camera with the reverse of the EOTF. flat encodes a linear captured range of brightness we assume). In principle, one can choose any desired Lm that we discover. peak brightness, also called peak white, and log- to more fully peak the gamma function or reference display screen. may use its brightness (for completeness, a person skilled in the art may What the display screen calls peak brightness is a display is the brightest color your screen can make, and that's three R, G, B, like 1023 for example. occurs when the channel is driven to their maximum and 0 is the same At the same time, a display screen around it is used to define the color gamut. creates a starting point) but, after much analysis, if only one value If it is necessary to use (Then, one can does not need to e-transmit the coding) to any recipient The specific reference used for coding is the peak white value and no If confusion cannot occur, we think 5000 nits is a very good practical value 28834.1284 we found). In such a scenario, one can use raw input brightnesses above 5000 nits. For example, within the reference brightness range of 0.0001-5000 nits, the final display screen and grade to visualize equivalent appearance brightnesses for wherever desired, if such data is displayed on a 20000 nits real display screen is to be displayed, level-up functions will optimize this appearance. Real display screens that are even brighter than the reference display screen for All its information can be added as meta-data. person main space Our reference to the color inside is outside the luminance-based primary color space for example, 25000 nits, this is used to artistically re-define scene colors. The first strategy is already inside a certain RGB space with existing teachings. (for example, two colors with the same primaries but different peak white placing the gamut side by side once re-normalized to [0,1] and one of these is typically our main luminance axis and any other coded with any of the available log-gamma functions R°G°B, defined by the space) how colors are determined should not be confused. At the same time, lower dynamic range display screen (for example, 1200 or 100 nits) or just for dark foggy scenes going to 50 nits graded images can be encoded in this main [0-5000] brightness range and then to our universal luma encoding which should not be scrambled conversion (for example, luma codes still have an allocation of either a constant or variable code allocation must occur via EOTF). We are talking about chromatic issues which separates dynamic range considerations from doubling and in both directions and, in particular, luma along the axis, for example, which luma allocation function should be used? We teach a luma-(u, v) chromaticity that is , but this encodes color Mapping for display screens of other peak brightness in space When done, re-allocation functions can also be used.

Optimal values of gamma of 2.4 and rho of 25 allow the person to perform all scenarios (eg. images). with at the same time some bright regions, middle-bright regions and dark to use a single luma allocation function for regions) If desired, after a few considerations and remarks, 28834.1284 especially if someone only wants to encode HDR (for example, although HDR metadata to derive an LDR view from pixel image coding. Although we can co-encode data as additional mapping functions, we In all applications, this optimal single log-gamma function The image resulting from the application directly has a good LDR. We don't need to get the view) this is done.

Several chronaticity representations can be used with our luma recognition approaches, but we found that the (u, v) space works especially well. We are also a references to selected white, for example: (u”,V,) = (u,v)-(u_D65,V_D65) We can define the defined (u', v,).

We assume that an input color contains information about brightness and chromaticity. When we state that such color is represented in such color representation We do not point out, but this information is derived mathematically and, for example, if the input If the image is represented in XYZS or some uniquely defined RGB, this is fulfilled. Although we are interested in the emerging HDR coding technology Although we emerge in a new way as a basis for defining useful luma, The expert reader can learn from our teachings, for example, how to calculate lumas from RGB coordinates. will understand how it can be defined as Y=al*R+a2*G+a3*B RGB to a defined desired white point compensation It is done by weighting the coordinates b where, al, a2, a3 are constants and depend on certain R, G, B priorities and a choice of white point. can still be elected by granting After that, these shines with our EOTF they are processed.

At the same time, the expert reader is aware of our final view of an image to be output. will understand how we can generate the information and we can use Rec_HDR (a desired We can call HDR (close reconstruction of the input) and this, for example, XYZ_output (if luma normalized [0.0,l.0], we use characters for luminance Y or also L and Ya or also v symbol for luma 28834.1284 (we use) can be done advantageously. But we also, for example, specific R”G°B” or other to drive a particular display screen Rec_HDRs to another color representation similar to device-dependent encoding. We can transform it. In addition, Rec_HDR is also an additional HDR image or we can have the desired view on a 1200 nit display screen directly to the driving RGB image needed to visualize We can map directly, or alternatively, a first Rec_HDR] ,e We can map (for example, for our 5000 nits reference display screen) and this, through an intermediate step, drives a 1200 nits real display screen It can be advantageously set as Rec_HDRZ to keep the

As new tools for this meta-data (rho, gamma and, if desired, Lm), we can undoubtedly at the same time, new data containing this meta-data, not related to prior art, image (although the pixel matrix encodes the color components Although whatever old strategy can be used for YCrCb, we hope In a MPEGHEVC structure we force the YuV coordinates, but for example DCT or The rest of the encoder (the decoder) that does the advance length encoding is us a LC or a piece of software that performs transformations according to our applications as long as we have it is ignored). In some HDR images, brighter pixels can also be several at the same time, for example a pair of lamps in a night scene it could be. In a mode-ii image, the relationship will be different again. Bright and dark some sufficient difference between regions will still exist (so that aHDR images (assuming a simple evaluation here) because this is just Because relatively simple functions can map to Rec_HDR not, but at the same time, even in the direct creation of LDR, the person is in some way This is because it may require a contrast appearance. But on the other hand, brightness Because LDR intervals can pull towards and into each other within a certain dimension gainut has its limitations. But what's important in all this is one's some signatures of whether the image is an LDR or HDR is visibility. Only mathematical image analysis algorithms encoded (for example, the final quality of the images for e.g. production 28834.1284 (real-time television production, where cost is less important) Not only can you analyze the range view, you can use an image analyzer for this. unit 177 can be used. But in general, our coding technologies A human color at the creation end in their highest quality format will be used with a rater and this human will typically see how the system behaves. as an HDR and LDR viewer (for example, LDR and HDR views are (what they look like) can see its rater keyboard can turn the dials and ultimately create an LDR that he or she is happy with. can encode image and HDR reconstruction functions. Typically buyers Note that there is no need to do a full analysis of the situation. Them, Is the normalized image they receive an HDR image or an LDR image? image and which LDR image is variable. They don't need to be careful. They are just the functions they take all it needs is what the functions define and/or the single image is what it defines. So typically the signal determines what type of signal it is. It will contain an indicator (lND) indicating

Only image with an optimal function Lm = 5000, rho = 25 and gamma = 2.4 signal can be specified for HDR-images (for example, there is only one HDR view and no LDR pixel matrix, just potentially an LDR view some color mapping functions to derive the image). lm = Instead of 5000, a content creator can set a higher Lm value (e.g. 10000 nits). or when it wants a lower one (e.g. 2000 or 1500), the method is the same can be used at any time.

One can also design additional gamma values and, for example, set a desired a gamma Rec that takes environmental visualization into account. 709 encoder gamma and The final gamma created from a gamma equivalent to gamma 2.4 can be determined by definition. Person skilled in the art, with equivalent gamma not the value of gamma in the formula, which also has a linear part, 28834.1284 but when starting in black as a gamma rather than a linear part, Rec. Gamma is the best approximation to OETF coded 709. He will understand.

In parameters rho and gamma a human on a 100 nits display screen to give a coded image that looks good according to the color evaluator. A method of encoding a high dynamic range image is further optimized where at least one of the parameters rho and gamma is preferably a human Optimized by the rater. Although some variability will exist, but depending on the scene, what appears to be a reasonable amount of time for an HDR image There is clearly a range of parameters that will give the LDR competitor and thus, this It becomes a method that can be positively defined and identified.

Due to the variability and complexity of images, the process generally a human rater selects the best looking LDR image and therefore It will determine the incoming rho and gamma parameters. However, at least partially automatic image analysis algorithms may come up with good looking values and Then, for example, typically, the rater just checks whether the generated LDR is actually is the embed he likes, all data (one or more views to one or more image pixels + functional meta- that define further color rendering data) before all data is written into an image signal encoding, just a must be confirmed by pressing the accept button. Darkness For curves that further affirm the images/scenes, one can shoot (for example, a night scene with no or perhaps few small lights, background moon-like in the background), the larger part of the codes is typically the larger in the scene will be made available for dark colors and this will be for example 2.55 (and a optimal rho can be chosen for this), e.g. one larger value of gamma may occur when selected. Brighter objects are proportionately brighter. For images that have a large amount of smaller with brightnesses much darker than the maximum object brightness smaller values of gaina, when only one pair of yainas exist, 28834.1284 for example 2-2.2 (2.15 is a good example) or even less than 2 and more than 17 A large value, such as 1.2, can be used.

Encoding a high dynamic range image correspondingly to the method An image coding device has several variants and includes the following: to obtain the pixel colors of an input high dynamic range image. entered and here, pixel colors have information about a brightness and a chromaticity. has; - to derive a luma code (v) of the brightness of a pixel color designed to implement an inverse of the mapping function. rating management unit (202), mapping function P-I) is predetermined to include a first partial function defined as and where rho is a tuning constant and v corresponds to a luminance to be encoded L = LmPy, which is a gamma transform of luma and a second part mapping is defined as where Lm is a predefined reference notation is a peak luminance of the screen, and gamma is a constant preferably equal to 2.4, - luma code and rho that can be connected to a Video memory or network and a color associated with metadata that includes at least one of the gaina parameter A pixel matrix image containing pixel colors encoded by the component A video transmission device designed to encode and transmit the image signal S_im. an encoder (210) connected to the connection (221). A typical variant may have a rating structure (to identify at least one HDR image) software that the rater implements, but now, its codings and/or graded views are combined to derive our specific with our technologies), but the device can also be inside a camera, which In this case, rho and gamma can either occur together or simultaneously, with one or two rotations. It can be changed with the knob. 211 typical of a Video transmission connection will understand how to put it into practice and that this can, for example, be a standardized Video cable output, for example in internet packages 28834.1284 a protocol for encapsulating, a protocol for writing onto a blu-ray disc protocolized hardware, etc.

Corresponding to the encoder, various decoders operate largely similar can be, but there are still some additional variables on both the encoder and decoder side. they can exist, as currently defined, once the code allocation When encoded, it must be uniquely understood by any recipient.

To decode a high dynamic range image encoding (S_im) An image decoding device comprising: - get the high dynamic range image encoding (S_im) and from there, process An encoding as defined in claim 1, containing the luma codes to be passed. method designed to derive an image encoding (Im_1) from receiving and formatting unit (388); - image encoded Tm_1 from a high dynamic range image (REC_HDR) designed to implement a color mapping strategy to derive a color mapping unit 305 and wherein the color mapping unit P-l is defined as containing a first partial function predefined mapping function in image coding (Im_1) pixel lumas are arranged to apply on v and here, rho is an adjustment is the constant and V is the Iuma code corresponding to the brightness to be coded and a second A predefined reference display screen of “Lm” for partial mapping where L is the peak brightness and is defined as LmPy and gamma is the high dynamic to obtain the brightness L of the pixels of the intermittent image (REC HDR). preferably a constant equal to 2.4”.

Of course, as a variant, it has a reference [0-5000] brightness and builds from there. instead of or in addition to the specified color space (e.g., XYZ), various code decoders also decode another image as output. 28834.1284 can hold. For example, outputting a Rec_HDR reconstruction of an HDR view In addition to an output for the LDR image, the decoding device provides a second to the output (or a display screen request as the output image) may have the same output (depending on the system being connected).

We have additionally invented some interesting applications such as: Encoding a high dynamic range image involves the following steps: method: - from an input high dynamic range image (HDR_ORIG) to a low dynamic a mapping function to derive the interval view (LDR_CONT). to determine where the high dynamic range image (HDR_ORIG) is A brightness-correlation of pixel L of a low dynamic range image (LDR_CONT) as a lumasina of a pixel Y = c * log10(a * Ll/y + b) + (1 is transformed through the application of a specified function, where, coefficients are specified so that the function lies in a range [0,1] for values L and Y The function is such that L=0 maps to Y=09 and L=1° maps to Y=1 is normalized and represents a certain behavior of the function in the middle of the Y-range. There is an additional constraint specified that applies near the value and Thus, the form of the function can be controlled by a single parameter a and - to a Video memory or network, the low dynamic range image (LDR_CONT) and one of the high dynamic range image (HDR_ORIG) and at least one forwarding to a video transmission link 221 linked to the parameter.

This well-chosen logarithmic function produces an optimal mapping. allows and after that, especially if one only needs 10 bits for the luma component (and for example, if the chromaticities have 8 bits for u and V) provided, the minimum can be uniformly quantized with visible error.

The resulting image is LDR CONT, o the brightness effect in various brightness sub-regions as a type of contrast-smoothed version of the HDR image, a 28834.1284 It can be called LDR image. If one uses the correct code allocation function If he chooses, the person can directly view the program on an LDR display screen. can use this LDR_CONT image for visualization, but this is not necessary for all applications of our invention, as some are HDR-only Can use LDR CONT as a pseudo medium for encoding.

In which the additional constraint lies at or near the middle of a Y-value or Y-range Application of the power function (LI/y), which is the inverse of the function of claim 1 The result obtained via L-value and in which Kl is a constant, for example It defines a functional relationship between a parameter such as L^l/y(Y=l/2)=K/a, A method of encoding a high dynamic range image is provided. Advantageous As a rule, one is on one side of the curve in those regions where the most interesting activity occurs. Evaluation example of a bright sunny outdoor and an indoor scene grading outdoor colors in M_HDR° to HDR medium gray may want. Interior colors with true typical interior brightness should not be created because we are in a typical low-light environment for television viewing. You are watching the movie in a different environment. So precisely, these are also Because they are not fully rendered, indoor colors may not be suitable for sunny outdoor should not be rendered at 1/100 of pixel brightness, reference pointer only It is not clear to any buyer what the reference basis rating will be. An exact copy is created. We take into account the adjusted average viewer view We need to add more details and this will make the interiors look realistic, especially in HDR view. It should not appear unnaturally dark. We call these colors "sunny outdoor spaces" 1/10 of the average brightness of the image area colors, around +- 100 nits We can rate. 28834.1284 In which a human color gradation will be transmitted to the video transmission link 221 encoding a high dynamic range image determines an optimal value A method is provided.

Preferably, on the creation side, our technology is at least minimal in critical areas, for example. to choose optimal a-values for artifact or for a good overall appearance. allows. The receiver, that is, an a-value of the encoded image and/or intended certain characteristics to correlate with the physical characteristics of the show's screen. you don't need to know the algorithm, but rather it just, which (reverse) Which functional form corresponds to the function's a-value, for example? you need to know.

In which the automatic image analysis unit 227, for example, calculates a a limiting brightness of a range of median or occurring brightnesses Characterizing the brightness of pixels in a high dynamic range image, such as a high dynamic range A method of encoding the image. At the same time, human rater can indicate where values of interest are in the image and, for example, 0 is a can make scratches on the image and then the unit 227 The 95% lying above 0.7 are mostly bright colors. can create.

There may be several pre-designed decision-making algorithms in the coding unit.

High dynamic range of chromatic coordinates (u, V) of color coding in CIE XYZ-coordinates of the colors of the pixels in the image (HDR_ORIG), a. ..l constants and preferably a:4, bîCîO, v : gX+hY+iZ e jX+kY+lZ, They are derived through fractional equations of type . 28834.1284 To encode a high dynamic range image, a image coding device: - from an input high dynamic range image (HDRIORIG) to a low dynamic a mapping function to derive the interval view (LDR_CONT). and a rating management unit (202) arranged to determine a brightness of a pixel of a high dynamic range image (HDR_ORIG). correlation L of a pixel of a low dynamic range image (LDR_CONT) lumasina Y = c * log10(a * Ll/y + b) + (of a function determined as 1 is transformed through the application of , where the coefficients are specified and thus, the function is on a range [0,1] for L and Y values L=O°in Y=07a and L=1° maps to Y=1”, the function is normalized and represents a certain behavior of the function in the middle of the Y-range. There is an additional constraint specified that applies near the value and Thus, the form of the function can be controlled by a single parameter a and, - low dynamic range that can be connected to a Video memory or network image (LDR_CONT) and high dynamic range image (HDR_OR1G) encoding an image signal S_im containing one and at least parameter a, and connected to a Video transmission link (221) designed to transmit an encoder (210).

A system that allows the human rater to choose a specific value of a. an image coding device including the user interface unit (203).

Therefore, a peak brightness of a display screen on which coding is done and/or brightness statistics of the high dynamic range image (HDR_ORIG) based on such parameters, a is a given An automatic image analysis unit (227) designed to determine the value An image coding device containing 28834.1284 Inside the gradation management unit (2029 high dynamic range The image (HDR_ORIG) is edited to determine its chromatic components. Due to an image coding device and luma-independence, this also low dynamic range encoding (LDR_CONT) of the original HDR input their chromaticity will be: a...l constants and, preferably, values: a=4, b=c=0, gX+IiY+iZ jX+kY+lZÄ An encoding of an image of pixel values and a request for An HDR image containing at least one value of the function parameter a”n1 coding signal.

Although some applications can transfer entire function definitions (e.g., a system that may not have pre-agreed function information) for the buyer or if the standard is agreed on only one curve, but the creation If the party wants to use another curve, then 0 is given to any receiving party. need to do signaling), if the functions of our applications If they are simple, like some others, only one or a few coefficients are involved in communication. presence may be sufficient to recreate their functional form.

such as a Blu-ray disc or a memory stick containing the HDR image encoding signal a memory product.

To decode a high dynamic range image encoding (S_im) A decoding device 301 comprising: - receive and process high dynamic range image encoding (S_im) It is designed to derive an image encoding (Im_1) to be passed from there. receiving and formatting unit (388); - extract a high dynamic range image Tm_1 from the input image To apply a color mapping strategy to derive (REC_HDR) 28834.1284 an arranged color mapping unit (305) and wherein the color mapping unit of pixels of a high dynamic range image (REC_HDR) Y = c * log10(a) of the mapping function to obtain the brightnesses L * Ll/y + b) + (1) in an inverse image coding (Im_1) pixel lumas onto Y and a, b, c, (1, and y for the image decoding device are known constants.

For the unit 603 which has such behavior for darker colours, there is no The tone mapping function satisfies what we need for invention, but above the log-basis equation there is a simple There is a pragmatic approach. For the lighter colors above the behavior is typically for example, further up the range taken by increased darker colors a function that maps light luminosities typically in a non-linear manner There will be a gentle compression with, Now, one may want other values too may have complex HDR images, but such extreme cases by the rater (or an automatically rating algorithm can be handled by means of a suitable random curve definition.

On the decoding side, the chain of processing consists of the transmitted LDR image(s). Significant inversion, being capable of calculating Rec_HDR It needs to be translatable. Significantly reversible statement Exactly the same color component in Rec_HDR as in our original M_HDR We do not have to obtain the values, but there is a tolerance for color differences. It means they must be within the limit. Therefore, the buyer ultimately Color conversion needed for upgrading to HDR-view Rec_HDR Must be capable of obtaining the functions and can do this from MIHDR to LDRýo (or LDR_i) was originally used on the receiver side when it did down inversion of rating functions and the form of these functions through obtaining information or when needed to upgrade to Rec_HDR. by calculating them by directly taking the inverse functions heard does. This situation gives an LDR-view to the rater's own preferences. for the random tone mapping function that it can define to fine-tune, 28834.1284 its normalized LDR and HDR in a manner understood by the skilled person. The need to define a monotonically increasing function regarding the It means he will hear it.

A high dynamic range image as claimed in claim 13 an image decoding device 301 for decoding the encoding S_im and - receiving and formatting unit (388) without high dynamic range image coding (S_im) is an a-parameter value that defines the format of the mapping function. and possibly also derive the inverse of the mapping function It is arranged to derive at least one parameter, such as a gamma value y.

The creation end can pass parameters and if only 1 parameter is sent, our Since some of our applications can specify other parameters, the code specific sub-regions of the luminance coverage of the HDR image to the gamut of the space Very useful to send a family of curves with different behaviors regarding allocation There will be an approach. For example, gamma may be fixed and pre-agreed and the person is placed somewhere within the image(s), or attached to it(s), or in a separate contact. path (e.g., a television station will use a certain a-value from now on) You can state that you use it and communicate this regularly) through It just sends an a-value.

As claimed in claim 13 or 14, a color is mapped therein. The unitary 305 has a universal color representation similar to a CIE XYZ space. mapping to the u and v component of the Tm_1 pixel colors of the input image a highly dynamic an image decoder to decode intermittent image encoding (S_im) device (301). 28834.1284 Preferably, we use a combination of colors in the chromatic direction as defined in the text. We implement brightness direction allocation with smart allocation and thus, quantized the total error of the colors (for example, deltaE2000) for use and for example at least the re-constructed REC_HDR is too large and even if its HDR signal does not translate to a true 10000 nits display screen The reference level of the reference display screen in which it is coded is 5000 nits. It may be a processed version of it. After that, the decoder is need to do the inverse of the color space mapping, and this typically Mapping slot colors to some universal color space like linear XYZ It will be implemented through.

A decoder as claimed in any of the above decoder claims a method for decoding high dynamic range image encoding (S_im). image decoding device and wherein the color mapping unit 305 is a high a second color to derive the dynamic range image (RECaHDR) additionally edited to implement the mapping strategy and this reference a brightness dynamic range lower or higher than the dynamic range is the image. In some HDR images, brighter pixels may appear at the same time. For example, in a night scene there might be a pair of lamps. In a mode-ii image, The relationship will be different again. Some sufficient difference between bright and dark regions will still exist (this is how aHDR images are created) assuming in the evaluation) because it's just relatively simple Not because the functions can map to REC_HDR, but because the same At the same time, even in direct rendering of LDR, the person has a somewhat contrasting appearance It's because he wants to. But on the other hand, brightness ranges are within a certain range. size, they can pull towards and into each other because the LDR gamut There are limitations. But what's important in all this is that one's image is You can see some signatures of whether it is LDR or HDR. Only mathematical image analysis algorithms are encoded in images (for example, the final quality of the images is less than the cost of production, for example) view of dynamic range (real-time television production where it is important) 28834.1284 But in general, our coding technologies A human color at the creation end in their highest quality format will be used with a rater and this human will typically see how the system behaves. as an HDR and LDR viewer (for example, LDR and HDR views are (what they look like) can see its rater keyboard can turn the dials and ultimately create an LDR that he or she is happy with. Can encode image and HDR reconstruction functions.

In some simpler systems, our technology is a single type of “closed” system. The available and intended optimal (reference) HDR display screen for e.g. It could be 5000 nits. However, for a 2000 nits display screen, for example, Additional functional instructions on how to map can exist and this will typically be done by starting from REC_HDR, but Also, LDR_CONT/Im_1 values are taken into account in the image It can be done differently using .

A decoder as claimed in any of the above decoder claims. a method for decoding high dynamic range image encoding (S_im). image decoding device 301 and wherein the receiving and formatting unit 388 is At least one reference display screen in which the input image is encoded as Im_1 get a peak brightness and possibly also a gamma value and from there it is arranged to derive the inverse of the mapping function.

Indirect ways to uniquely identify a code allocation function exists and, for example, the person is at the top of the intended (reference) display screen. About a set of functions to use for ranges of brightness may agree. After that, another true demonstration with another peak brilliance screen to make it look optimal for its characteristics. It can additionally map REC_HDR, but at least 0 which code It is necessary to know that the definition is used. 28834.1284 The above decoder claims 13 through 16, incl. a high dynamic range as required in any of the an image decoding device for decoding the image encoding (S_im) (301) and here, the receiving and formatting unit (388) extracts Im_1 from the inputted image. color to derive high dynamic range image (REC HDR) pre-agreed data required to be used by the mapping unit 305 which specifies which of a set of remaining inverse mapping functions It is arranged to receive a code such as an index number.

The actual coding and transmission can be done in various ways, for example just 3 different curves a standard that allows this program to transmit this or a program A “2” part of the function is used.

As claimed in any of the above decoder claims a display screen including the image decoding device.

How to decode an acquired low dynamic range image (LDR_CONT) A method that includes: - receiving and processing a high dynamic range image encoding (S_im) Deriving an image encoding (Im_1) to be passed from there and - extract a high dynamic range image Im_1 from the input image Applying a color mapping strategy to derive (REC_HDR) color mapping via and here, the color mapping unit is high obtain the brightness L of the pixels of the dynamic range image (RECIHDR). Y = c * log10(a * Ll/y + b) + (1 a) to apply the inverse on pixel lumas Y in image coding (lm_l). and a, b, c, d, and y are known constants for the image decoding device.

As suggested in Istein 209, a captured low dynamic range image (LDR_CONT) is a method of decoding the image and receiving a 28834.1284 unique inverse of the mapping function Y = c * loglO (a * Ll/y + b) + d It includes the receipt of any information that identifies you as.

The invention has been implemented in signals for various purposes such as defining parameters. similar to intermediate cases containing the core technical requirements of applications It can be achieved in many (partial) ways and its many applications are in communication. by various means such as finding, using, color processing etc. various possible signals and various paths various hardware combine components or use them in various consumer or professional systems. methods can be used.

BRIEF DESCRIPTION OF THE FIGURES These and other aspects of the method and device in accordance with the invention are discussed hereinafter. by reference to identified implementations and applications, and specific, non-limiting representations that merely exemplify the more general concept made clear by reference to the accompanying drawings which serve as will come and be evaluated.

Figure 1 shows an LDR corresponding to a main HDR rating Mandatory by the content creator to create ratings Our generic with a gloss of an object to be represented will be used To associate what we will call as a luma code An example of a family of such available luma code allocation curves shows schematically (HDR grading typically occurs on a RED camera with a high quality camera - either celluloid film or an HDR or typical upgraded to - after camera capture, human-like appearance fine-tuning, however, one can also adjust the brightness in grading pseudo-HDR with enhancement or special effect from an LDR camera capture processing can create a human colorant such as address physical limitations of the camera, such as reduced saturation 28834.1284 can enhance colors, but can also enhance the artistic to improve its appearance, for example through dark corridors and likingma can improve this and this main grading is quite complex can be done and should be done with caution); Figure 2 shows how to grade and analyze HDR image(s) in accordance with the present invention. schematically shows an implementation of possible devices for coding; Figure 3 shows some steps to use the encoded image(s) according to the invention. shows possible devices schematically; Figure 4 shows different codes from the set depending on the brightness characteristics of the image. A number of allocation functions (often 3 is sufficient) to determine how one shows schematically what it can choose; Figure 5 shows a picture of any of our possible luma codes. shows schematically how one can identify the chromatic component; Figure 6 shows specific visualization such as a typical average room ambient brightness specific requirements of their systems and how our functions currently sennatically some additional notational applications of what you can combine shows; Figure 7 shows some exemplary mathematical equations for such functions. shows definitions schematically; Figure 8: When a code step is made, what happens throughout the code range? The steps of the visualized luminosities DY/Y are shown in the selected curves of Figure 6. schematically shows some equivalents for , where luma When a typical value of 10 bits is used for codes, in this example, this color Quantization errors are only related to noticeable differences (JND); Figure 9 schematically shows one possible coding system with encoder and decoder. and in the case of our applications, `these are in S_im, The pixelated image is an HDR image, or more accurately As a rule, when used directly, it mostly uses HDR-like They are used in such an approach that it has the appearance; Figure 10 schematically shows one possible coding system where, in the case of our applications, they are pixelated in S_im 28834.1284 the image has a more LDR-like appearance or It is true that it often has an LDR-like appearance and, for example, LDR of peak brightness around 100 nits, such as 500 nits peak brightness for significant direct visualization on display screens are used in such an approach as is appropriate and Figure 11 shows an HDR image and parametric analysis from the HDR image. a derivable corresponding LDR image in an image signal. Shows possible coding schematically.

DETAILED DESCRIPTION OF THE FIGURES High dynamic range (HDR_) images/video typically currently They have a different brightness distribution than the used images/video. In particular, the dot-to-average brightness of high dynamic range image data The ratio can often be much higher because reflective objects in the room They may have relatively darker colors and then, lamps or light There are a couple of very bright objects as effects. LDR images typically A more or less single (most (at least in important parts of the scene) while HDR imaging technology deals with the real world and at the same time this objects in bright spotlights and others in dark corridors has scenes with highly variable illumination, with shadows it could be. But on the visualization side, this also means that one's HDR image re-defining the appearance through color mapping thus, its LDR is better suited for cysteines and this determines what our LDR rating or rating is. It means that we will determine the nominator as . At the same time, only one When the HDR image is encoded, the statistics of the brightnesses are subject to various LDR coding The luminance code allocations known from the technologies are now good with the gamma-2.2 type. It doesn't match that way. 28834.1284 In the unrestricted implementation of Figure 2, we have a rater has already prepared a master rating of HDR_ORIG We assume, for example, that this is a 3x16 bit linear XYZ image. we assume, but in the following, we first determine that the pixel color coding is a brightness We focus on the correlation (for example, a luma code or luminance value) and we assume that this value will be a representation [0,1] (the skilled person would, for example, 0...1024 encoding how to do alternative implementations). We, grading and coding functionality in one grading device We assume 201, but they could also be separate devices (essentially, we just teach a coding unit like part of an IC7).

A user interface unit 203 includes a human rater (all units such as on a dedicated IC or a generic processor rating control (may be running software) by user input USRINP) and, for example, a combination of indicators to change values. can be connected to the keyboard and, in particular, the a-value of our curves below or rho and gamma values and even Lm value can be selected. Rate it management unit 202 sends a mapping call as described below. to determine as well as according to the rater, for example, VCl traditionally with an AVC-like MPEG-type encoding LDR_CONT is about to reach the optimal LDR output that can be encoded. It is edited to apply HDR_ORIG to an input HDR or images.

The rater performs its own monitoring on connected calibrated monitors. you can look at the results typically and, for example, if an LDR view at 0 S_im If it specifies an HDR encoding as directly to the LDR image (decoded without encoding or even, still DCT just the recolored view before encoding) and simultaneous LDR on a reference HDR monitor, typically 5000 nits white LDR_CONT can look at the Rec_HDR image that can be recovered from the image.

The rater uses a brightness-correlation mapping function (this is the same At the same time, in principle there may be mapping between R_HDR and R_LDR) FHZL can determine, through it, the brightness of the HDR image (or 28834.1284 lumas) are converted to the luma values of the LDR image (or conversely, these the function will typically be reversible, the inverse function FL2H It can be used to reconstruct Rec_HDR° from LDR_CONT and is typically As such, such an amplification function is stored in FL2H S_im). our coderism In other implementations, this can be done automatically, for example by a single hardcode by allocation curve (for example, half way in this family of curves) or a sequence of images or successive images. analyzing the sequence (for example, a shot, scene, or even the entire program) through and, for example, median or a weighted average brightness or at least one sub-range to select a curve through a set of rules This can be done by using a count of luma occurrences in: When median<:Y, then curve number Z is used. This mapping function thus, one can access an encoding or Conversely, an optimal HDR image can be obtained from an encoded LDR image Once you have it as an optimal tool, similar to what the content creator prefers Allows you to derive the LDR image that appears as At the same time, reverse can be readily determined (our brightness mappings typically because they are reversible) and this e.g. The evaluation can be done at the creation side or at a receiver side. and this function can then be provided with an LDR encoding LDR_CONT HDR_ORIG is a approximation of the original base grading on the basis approximation (after quantization and DCT approximation effects) Allows reinvention. For storage or transmission to a receiving end, the person The HDR1 to LDR7 mapping function is FL2H, its inverse is FH2L, LDRICONT of the LDR image and HDRIORIG of the corresponding HDR image (and, for example, after an economical mathematical transformation, these images any combination of any close approximation) can code. However, for bit budget reasons, you only need to view one of the images. It means to store/transmit. We encoded the static LDR_CONT of 210 or lm_lDR first images defining All_l textures of movie objects We assume that it encodes and formats (for example, classical MPEG encoding- 28834.1284 etc. DCT making and run length coding). He is also chosen formats the FR2R prefix of certain mapping functions singular], for example, into a toplain image signal S_im FR2R=FL2H (expert The image can be edited with artificial intelligence to analyze regions in the image and which of the regions have certain problems in HDR encoding, especially in mode-ii type It can be seen that they can give. In particular, regions prone to 0 banding can identify and recognize adequately textured regions and thus, they are displayed with a lesser amount of luma and/or color component codes. They can be coded. In some applications, this unit may replace the human rater. without, a final encoding may occur automatically). After that, encoder 210 transfers this Im_l and meta-data to a blu-ray disc or solid state memory device stores or (near) real-life data in a memory 299 such as a similar memory product. in a television studio for a time-streamed DVB-T television signal. If rating occurs, some network technologies transmit the signal over 211.

We show one possible Video transmission connection (221) and this is a master BD a bus or cable to the disk or on a content-provider-owned server. It may be temporary memory storage, but images can be stored via various technical systems. Several such connections may exist to output signal(s) and, for example, antenna also has a second such S_im output connection (not shown) it could be.

We typically follow a series of standardized scenarios (at least if a If not single mapping, then a first step in a succession of brightness and typically also color saturation mapping-like mappings) HDR brightness-correlation (for simplicity, we just we will assume brightnesses, but they are doubtless also, such (they can be any encoding of brightness) to go to “LDR” lumas Code allocation function(s) for OETF (or conversely EOTF) to be applied preferentially We discovered a couple of variants of what happened. Advantageously, these functions For example, they can perform the following, which is not necessarily mandatory: 28834.1284 l. The effect of applying curves is a change in brightness (brightness is a physical visualized brightness has a psycho-visual effect). 2. Brightness change in two directions (both lower and higher brightness) should be possible and, preferably, no information/details should not be lost or very little should be lost (e.g. curves should be at least must be invertible in a continuous color space). 3. The images resulting from the application of curves are perceptually They must be satisfactory and, for example, the human color grader Must be capable of making pleasant or relatively pleasant images (especially in perceptually relevant brightness ranges, contrast ratios are They must be properly protected). Especially since most of our applications start from an existing HDR Therefore, the functions display the encoded LDR-looking pixelated image. They can map HDR into a near reconstruction Rec_HDR. But Of course, only generically within any particular approach may not be possible and, for example, a specific technical coding chain is needed.

Now, one thing that is important to understand is that with our mod-i (HDR-view) system, is that the rater can identify such mappings arbitrarily because We just use the LDR-view image (without any re-construction, if can be done in a data-destroying way if 0 is desired by the rater) We need to derive, because in the coding approach we have, HDR-view image S_im in the image signal is currently single image It is coded as . However, in mode-ii systems, we have a binary criterion We need to fulfill: on the one hand, we provide Rec_HDR with good quality We need to be capable of reconstructing the image, but other On the other hand, most, if not all, LDR-views can be viewed by a rater. We want enough freedom to create however we want. 28834.1284 But overall, our coding technologies are their highest quality format will be used with a human color grader at the creation end and This person typically sees how the system behaves in an HDR and LDR renderer. (for example, what LDR and HDR views look like in reality) view, rotate the dials of its rater keyboard, and As a result, he created an LDR image and HDR reconstruction that he was happy with. can code functions. Typically buyers will have a full analysis of the situation. Please note that you do not need to do this. They, they get normalized whether the image is an HDR image or an LDR image and which Need to be careful as the LDR image is variable They are not. They just “blindly” implement the functions they receive. they need. Typically all they need to know is what the functions define and/or what the single image defines.

So typically, the signal is an indicator of what type of signal it is. (IND) will be included.

LDR brightness to reduce dynamic range in tone mapping correlation and HDR brightness-correlation our The part of the logarithmic curve that we can apply is from a general first form v = c * loglO (a * x + b) + (1 can be initialized where x is “linear” normalized to the range 0.1 is the input value and v is also normalized to the range O..1 to O..1 are used and are given by the following expression If we have HDR brightness on the x-axis, then our dark areas where we need to brighten or compress, for example on the y-axis We use logarithmic format to obtain LDR_CONT luma values. It should be obvious that it is necessary. 28834.1284 We apply constraints to additionally determine the curves. first two They are given by the constraint normalized range 0..1 where, input The values 0 and 1 are mapped to analogous output values and, for example, V becomes zero. When is equal to one, also X must be 0, and when V is equal to one, also Then x must be equal to 1*: 0 : 10 6 _b from here it follows b : 10-d/C -__-b . . .. l _2 _2 1 : V this can be written as 10c >i< 10 c - 10 c = a Finally, in the middle of the log scale, at v = 1/2°, we find the function with a (providing a linear luminance change at this position When a is varied) 1/2-d We apply the quixotic expression that should be 10 C _ b = K› where K is a constant and is reconstructed as 1025 * 10 C - 10 C = K. can be written. By combining the next two constraints, we get 10 C We eliminate the term and obtain K * (106 _ 1) = a * (1026 _ 1)' and we do this with the following solution by substituting 3/ : IOC› We solve as K * (V _ 1) = a * (Ji _ 1) giving: K2-2*K*a+a2 The solution is valid for a > 2 * K. By choosing a value for K, then The curves are determined by a single parameter a, which depends on a light sensitivity. or film speed parameter and therefore, we use a as our 28834.1284 We call it the exposure index parameter for our curves. for K The value we choose is as follows, K=8*\/î because this value of K° does not correspond to practically used exposure index values. Resulting in values for approximately the corresponding

Table 1 shows C-code implementations of log-curves and their inverses. are given and here, the variable names are used in the above equations They correspond to what happened.

Table 1. ANSI C code of proposed log functions and their inverse their implementation. static float LintoLog(float x, Iloat a) float K,b,c,d; K = 8*sqrtf(2); b = powf( 10,-d/c); back c*log10f(a*x+b)+d; static plot L0gt0Lin(I10at v, float a) float K,b,c,d; K = 8*sqrtf(2); b : powf(10,-d/c); back (powf(l 0,(v-d)/c)-b)/a; 28834.1284 Figure 15 shows a few exemplary curves of a proposed log curve family, It is plotted as follows, where this step size is log It can be easily observed at the middle position of the scale (value 0.5).

So now we have a set of functions that can be controlled by a parameter a We have the set. The grader is activated, for example, by turning a knob. For example, you can easily determine the optimal a-value for images for a shooting/operating situation. can derive and an automatic image analysis algorithm an optimal a-value can choose in a similar way. And we also encode an image or Sym To store the rater-selective a-value in the video signal, for example ilot or int (since we don't need many values, we use a-values We can encode it as A*a+B and thus, different a-values, for example 8-bit code Defining a data type whose word values can be allocated We can easily code this relationship using . So, for example, a LUT-like As an alternative to encoding the full curve in the signal, our technology implementations can only encode the a-value (once or twice, data with the same value for safety against degradation or for adaptability with different a-values) and after that, if used functions are run Not pre-agreed at the time, but agreed within the standard If so, which real function does the receiving end relate to the a-value? He will know immediately that he is associated. At a buying end, this value is then, lm_1DR converts the acquired image into a final image to be visualized on a particular TV. Can be used for mapping to image.

An even better way to propose a family of log-gamma curves: 28834.1284 and where, L is the luminance in Cd/m2, V is the electrical value and Lm is the peak brightness value of the display screen in cd/in2. of the person only if it can define a single native HDR code allocation curve, The proposed values of the optimal constants p and y° are p : 25 and y : 2.4.

We have at least approximately this range over 0 EOTF with inverse OETF functions we may regard behavior as corresponding to a higher degree of truth (but the person classically depending on how OETFis are defined, for example a low By making the brightness part linear, slightly adjust it to be more harmonious. may vary), for example: and where is a reference normalized by the reference white level Proportional to the indirect light intensity to be detected by the camera color channel R, G, B is a voltage and, for example, contact photo-electrons with R, G and B pixels are the linear voltages resulting from filling the buckets and E" The resulting non-linear signal can be predicted to be, for example, the luma code. And If we compare to the first variable, we get rhoiyu a as follows We can identify: 28834.1284 Certain optimal luma allocation functions above are used to generate OETF. additional justification: Current television systems have an end-to-end (optical-to-optical) linear It has non-transfer characteristics. This transfer characteristic is typical of low-light environments. provides accurate visualization intent for the television viewing environment; for example, "The Reproduction of Colour" by R.W.G. Hunt (Sixth ed., Wiley, 2006) publication 11.9, See sections 19.13 and 23.14.

Philips high peak brightness display screens (specifically, 5 000 In experiments with a display screen Philips with a peak brightness of cd/m2 end-to-end for future high dynamic range television systems television system transfer characteristics have been examined and the current end-to-end transfer characteristics will also be applicable to these future systems. has found its applicability. The narrative for this observation is transferred characteristic is determined by the television viewing environment and high For dynamic range television it will be the same as for existing television.

End-to-end transfer characteristics recommended for existing television systems OETF° (Rec. ITU-R BT. and EOTF° (Rec.ITU- R BT.1886) are determined by connecting them together. For example, Rec. ITU-R BT.709 OETF is issued through: 28834.1284 This OETFs Rec. Combining ITU-R BT.18867 with gamma 2.4 EOTF This results in an end-to-end transfer characteristic of: (HM9UVß-onanm arizrzoow Philips offered high dynamic range television using EOTF It offers to fully preserve the end-to-end transfer characteristics for its systems.

This EOTF has the normalized format: Function x _ (î) and Rec. A gamma 2.4 EOTF according to ITU-R BT.1886 can be seen connecting to each other. Therefore, preserving the end-to-end characteristic The OETF used with the proposed EOTF is currently recommended reverse of OETF (Rec. ITU-R BT., c The function must be interconnected and this is: This interconnection follows OETF (as an example, Rec. ITU-R BT.709 V:%&2%E222 ßrQMS>L20 28834.1284 To fill the proposed value of 25” for p, OETF additionally It can be simplified to the following: For V IZLZOOIS As equivalent, Rec. The proposed OETF for ITU-R 2020 is as follows: and where E is a reference normalized by the reference white level Proportional to the indirect light intensity to be detected by the camera color channel R, G, B is a voltage; E' is the resulting non-linear signal. And here: The simplest way is to derive the complementary image (for example, if HDRIORIG encoded with [N-bit color coordinate] on the disc, making any TV a significant LDR_CONT display for driving with extremely low peak brightness The co-encoded a-value and our chosen log-gamma function are only available through its implementation) is just to implement the mapping, can derive intermediate ratings for visualization. 28834.1284 Our invention can be applied in several ways in several applications. For example, if The rater tries to select an optimal curve for the particular image/Video at hand. If not, a default curve is chosen, for example with a=1100 (and if the a-value data type signal has no values to fill in, the receive end defaults to this will use the value. But otherwise, the grader would, for example, evaluate a curve a = 550. you may find that it gives better results, and then you can set the a-value on the disk. can write this value in at least one copy of the data type. If there are more copies, Typically, an associated presentation such as a timestamp belongs to this curve. reference data specifying part of a set of images, such as the same Additional specification data will be available from time to time.

Our methods use the image (Im_1DR) texture coding format. can be used, but for example, 0 is classically used in some applications of image encoding need less quality) with 10-bit luma encodings and even 8-bit It can work well with luma codes.

They can, for example, have a peak brightness of up to 10000 nits and Demonstrators may be between or around them. After that, although Although the display may still be old and simple, there is no HDMI or other connection to them. a future HDR content-delivery set-top box or A new high complexity decoding and color mapping IC in the computer can be serviced by and the set-top box we found and described offers any combination of options. of an old LDR image It will need some optimization between intra-object and inter-object. Please note. We want to see the internal textures of objects clearly, and still, An impression of the LDR image is a high-contrast view of the original scene. and, for example, the difference between a region of high and low brightness LDR It may not be perfectly reproducible with the image, but human to the extent possible that can be conveyed by the rater in the LDR. 28834.1284 still a remnant of this which makes lighting changes that are still optimal should be.

The rater or automatic algorithm may also be coded (e.g., type of scene captured) based on the characteristics of the image may decide to take a particular luma code allocation curve from the family. If e.g. 0 only dark areas (or just one small bright lamp) contains some light on the bright end for an increase in sensitivity for darker colours. may consider using a curve that sacrifices quantization precision. Some In HDR images, brighter pixels may also appear several times at a time, for example at night. There may be a pair of lamps in the scene. In a mode-ii image, the relationship is different again It will happen. Some sufficient difference will still exist between bright and dark regions. (A simple review here is how aHDR images are created.) (assuming) because this only applies relatively simple functions to Rec_HDR not only because it can do mapping, but also because LDR directly Even in rendering, one may want a contrasting appearance in some way. is indirect. But on the other hand, brightness ranges are interconnected within a certain dimension. They can be true and intertwined because the LDR gamut has its limits. But What matters in all this is whether the person's image is an LDR or HDR. You can see some signatures of what it is. Mathematical image analysis only algorithms are encoded in the images (for example, the final real-time, where quality is less important than, for example, production cost television production) can not only analyze the dynamic range display, Figure 4 gives an example of such an application. The Luma allocation curve is now, a logarithmic and power, with the gamma preferably not equal to 1.0,& is a combination of the function: 28834.1284 Note that the Luma codes (E horizontal axis) are scaled to [0,1]7 and a the corresponding brightnesses on the corresponding reference display screen. Note that the stops (y-axis) are given logarithmically. For example, if If one wants a single curve for a peak brightness of 5000 cd/m2, one 7:2.35si (for different gamma values, a...d coefficients typically differ will be) can use and derive.

In this example, we are using HDR images with a reference luma allocation curve of 401. assuming we can code and this is the same image with a sunny outdoor Simultaneously a darker interior-like, all possible HDR It will work fine on images. Now, if only bright lamps a pair (only those who need to visualize brightly and not be precise) a program playing in a dark basement with (that you don't need) or If we own the scene, we can use better sensitivity for such dark areas. behaving better, for example in the dark sub-range of HDR brightness/X-axis We choose another slightly different curve that has more provideable code. We may want to. Egri 403 would be suitable in such a situation. At the same time, other The scenario occurs where there are many bright sunny outdoor pixels. and since human vision is adapted to large bright images Therefore, it may not need to be quantized with absolute precision. Such a scenario, for example, when a movie is outdoors in Thailand and a temple is when the inside of the piece can be seen (the visualizer shows these dark interiors of the TV that he could decide to bring some light to the light, thus making what was ours a reality. (Note that we would want it reasonably encoded) scenario may occur. In this case, the human rater or coder device/algorithm for LDR_CONT used to encode these HDR images may decide that curve 402 is a better curve. 28834.1284 Figure 6 shows the same total log-gamma EOTF rationale, but now, a The intended monitoring environment is included and Figure 7 shows how these functions are implemented. Shows two examples that define what will be calculated. The part with “LDR” has been criticized are luma(v) code values and for example, for just an HDR reconstruction, we Typically, only our two-part (for example, rho part and Then we apply our exponential function (gamma part). Now, “gamma 2.4 is changed into a partial pre-gamma mapping chain, but now, the same It also includes a factor that takes into account low ambient brightness and is shown at the bottom of Figure 7. partial upper gamma submapping a 2.4 and rec. 709 OETF: v = 4.5L if L < Determining luma from brightnesses) is summarized as an equivalent value. LC(a) is the first part of the mapping and, for example, 2.4 is the rho-part without gamma.

Arrows alternate between an up arrow and a level up type transform, for example dark typically stretches from bright objects to mid-gray and, conversely, away from bright objects. shows the transformation. Q rec. A normal 10-bit quantized LDR according to 709 becomes an image. Second Rec. 709 downleveling correctly specified input, According to what happens to our main EOTF, which is expected as input reformats (redistributes accurately along the luina axis).

The result of the upper and lower chain of Figure 7 is then typically followed by a reference an actual display determined according to a gamma of, for example, 2.4, delivered to the monitor. will be sent to the screen.

Figure 8, Figure 6 functions DY”/Y, “when converted to dark object Shows a zoom in on colors.

Figure 3 shows a consumer's home or professional environment, such as a digital cinema. Shows an example of a live receiver-side system at its location. It's just Useful to have a single master EOTF for HDR image encoding not, but also, if we have different classes of HDR images If so, gamma would be very useful. In reality, if we If we do not decide to implement a single optimal function that is One 28834.1284 format decoder 388 0 depending on the format in which it was recorded/transmitted and received unpacks the signal S_im and decodes it. We are this non-limiting In the example, acquisition and initial processing may require some image manipulation. Device 301 (a set-top box, blu-ray player, personal computer, etc.) We assume that it is done, it is created correctly - the specifics of the display screen and possibly as the environment he would want - the final image to be visualized is a is transmitted to the display screen 302 (in this example, no additional color optimizer display screen without capabilities, but only with some hardware determined colorimetric characteristics). But undoubtedly, if this representation If its screen is smarter and, for example, a television, it itself is Most or all of the functions defined in the switching device 301 can realize. We can only provide all the needed shine we only have 10 bits and at the same time we can view all other images of different brightness. We need to adequately code the regions. But it's a trick can be used here, if the image having brightnesses adjacent to some codes can be borrowed from regions, and especially if the visual quality of these regions decays slightly by removing a few codes from their code range (typically through a simple operation of accepting the evaluative result and if another effort is to be tried, disagreeing, judgment and the taping is still too high for the original taped area. The more aggressive the case, the adjacent region may decay more to the gala or if the grader detects that the adjacent area has begun to degrade too much (may be slightly less offensive) this trick can be used. codes A simple approach to redistribute the local function part is linear or It is a non-linear modification.

In some embodiments, the image handling device 301 may also, or even alone, an LDR display 301 (display) for a second LDR display screen 330 LDR_CONT can be the image directly or dynamic range or tracking without setting the environment, its just a colorimetric conversion to RGB can be, but it is also an LDR_CONT derived from the encoded image. 28834.1284 The second optimally can be the compiler image) and this antenna Streamed over a wireless image/video/data connection via 399 is kept, but the core part of our invention is also just an HDR can be used to create images.

A color mapping unit 305 is, for example, an LDR_CONT encoded from S_im. receives the image and reads the FL2H mapping function from the signal or reading a mapping function F H2L and its internal This is done through conversion to the inverse FL2H mapping function. Converts to Rec_HDR7.

The constructability of many such devices or systems from our invention must be understood and 0 on professional or consumer cameras, display screens of any type (eg. may reside in a portable device like a mobile phone), in color grading software, Video enhancement devices, Video management audio and video systems, such as broadcast display screens in supermarkets It may lie in converters.

Now, one thing that is important to understand is that with our (HDR-view) system, is that the rater can identify such mappings arbitrarily because We just use the LDR-view image (without any re-construction, if can be done in a data-destroying way if 0 is desired by the rater) We need to derive, because in the coding approach we have, HDR-view image S_im in the image signal is currently single image It is coded as . However, in mode-ii systems, we have a binary criterion We need to fulfill: on the one hand, we provide Rec_HDR with good quality We need to be capable of reconstructing the image, but other On the other hand, most, if not all, LDR-views can be viewed by a rater. We want enough freedom to create however we want. 28834.1284 The inventor already divides one's coding space into a Y direction and a line of color planes. It has been previously reported that it can be separated into chromatic aspect (see unpublished reference (see document EP12187572) has been recognized (in any case this space not necessarily used to perform image color processing, but only can be used as an intermediate "value-holder" and whatever the color coordinates are, if we just restore the original XYZ_16bit of the master HDR via back mapping. If we can get it back, that might be enough). We see this schematically in Figure 5.

The gamut of all encodable RGB colors is 502 (and we have an xyY-like space We may want to use so that the top of the gamut, say 5000 nits, is has a defined maximum brightness through recognition of its white point. (all possible physically occurring colors can be encoded) The triangle xy and the luma are determined via the Y-axis 501. And more specifically, this The Y-axis defines the lumas Y and the corresponding luminances assigned to its code It is used by the function. To achieve a true shine appearance other than tone mapping, typically a color different from the encoding gamut to make chromatic adaptations to optimize for gamut May include edited. This unit is for example a 1300 nits wide gamut displayer. can apply the optimization (tunability) needed. After that, if This stretched region between the stretchable. The same thing will occur with other chromaticities and, for example, Document EP12187572 describes an implementation of how to do this and here, the chosen logarithmic functional form is followed, but now this [0,1] Y- possible Y° in the gamut for a given chromaticity (x,y) It is done through the scope of the . Now we want to make the same bushel in the chromatic direction. we want. Although desirable, the universal xy-chromaticity is in the blue region (B). has relatively small MacAdam ellipses and large ones in the green region (G) It is known to be. This is a big way to make a coding mistake (using (x,y)_l in the green area). Changing it to (x,y)_2 will not have such a big noticeable effect) means, but the quantization in the blue area is more readily noticeable.

Being competent to code everywhere with minor errors, we ultimately 28834.1284 Make sure that quantized coding distributes quantization errors more evenly. We want to be. An error function over the triangle (a color path 503 in the chart (shown schematically in one dimension) through which we determine we want and our chromaticity allocation function (semantically at 504 path H defined as a function of the same coordinates uv as shown also a function of x and yi can be defined as G) We can change this, for example, if an ellipse is part of the color space. If it is too long for the region, we can similarly stretch the color space in the 0 region and this becomes the inverse equivalent of compressing the ellipse. This redefinition is a high can be a highly non-linear function, but preferably, we encode We identify with someone who is simple. Namely, person ellipses are created from a perspective transformation. can deform: In these equations, x,y and u,v are color triangles and a...1 are constants.

A person can see that it is UVW from the color space defined by XYZ vectors. Corresponds to a fundamental transformation in 3D into the color space defined by vectors You can prove mathematically that it comes from a linear transformation matrix. It can be defined by .

We consider it to be a prior art mapping with reasonable uniformity (but Not to be used for HDR encoding (never studied) We have found it useful for describing chromaticity: X+15Y+3Z X+15Y+3Z 28834.1284 Here, the chromaticity plane coordinates are linear XYZ color space are determined directly from their coordinates, and thus, in that space we we can grade and then go directly to color coding We can move forward.

The next thing we will do is use the logarithmic luma code allocation functions above. is that we use and the Y (non-linear luma) part of the Yuv color definition use it as the definition and for the chromatic component of colors. is to use the definition above. This is a person's now classic LDR color It's about how HDR defines its colors as opposed to its coding schemes.

However, all these colors a person normally displays in the LDR (specifically, With Y,uv PAL-type YUV or YCrCb as used in MPEG29 which will not be confused) will be in neklenti (e.g. u and V” as well as, Since 10-bit for “Y”) has values in the MATHEMATICAL range, Two LDR coding technology and additional image coding chain (quantization, DCT- lemma) can do. Upon mapping back on the receiver side, we get a true HDR We get the image and it cannot be encoded in any other way, through this old technology.

Of course, color spaces can be designed following similar technical principles. and here the gamut cover is luma as long as one defines them as standard is defined through the new definition of the axis and thus, the image receiving end Can recover original master HDR.

Figure 9 shows a first possible useful image dominated by an HDR view. shows the coding system schematically and, for example, we still relatively We will transfer an LDR_CONT image with large dynamic range (for example, when displayed directly on an LDR display screen, o have areas that are too dark to be adequately recognized, for example 28834.1284 maybe, but still a great HDR image, if connected For example, a 5000 nit HDR display screen is available for HDR display over 958. can be). Our EOTF is particularly suitable for this scenario and then, parameters rho = 25 and gamma = 2.4 by parameter embedding unit 908 Sim can be embedded in the signal.

From our master HDR rating we use HDR_ORIG (here ref. no. 901, a The hardware unit is tilted but it is an image) we start. Color conversion unit 902 can do a color transformation and, for example, if the original has a relatively wide is left with saturated chromaticities within the color gamut (e.g., film as may occur with some colorants in the material) and, for example, typical Rec. Only one service of the consumer display screens of the 709 primes can be viewed. If placed in front of you, this color conversion unit 902 Rec. A pre-color to the 709 gamut It can already do gamut mapping. A dynamic range converter Interface unit 903 to obtain an LDR view image 905 typically some artistic guidance from a grader through implements functions. This LDR image can be reversed relatively simply can be achieved through color mapping, but it is also more complex and irreversible (data destruction, for example, from 0 images alone master HDR mappings cannot be reconstructed perfectly can be obtained via . Now, HDR is, for example, the inverse of the EOTFi of Claim 1 by color coder 906 according to our applications using is mapped into our Yiu”v“. After that, regular video encoding the encoder is made by 907 and this is a HEVC encoder or any similar It could be something. Finally, in these exemplary applications, our parameters of our colorimetric coding functions (at least rho, gamma and For example, formatted in a DVB or HDR television signal They are embedded as meta-data within the S_im to be transmitted as ATSC. a buyer On the side, a receiver will decode regular video with decoder 951 and but here, the decoder also produces a Rec_HDR image at XYZ for example. Our technology will be used to obtain 952. second color 28834.1284 After a second gamut adaptation via mapper 959, Take into account that the connected monitor has wide gamut capabilities. We can add. Display screen adjustment unit 957 additional display screen you can make adjustments and, for example, have a 2400 nit display screen An optimal brightness setting for the connected display screen is required. We use our second mapping functions to derive the view. and can also implement monitoring environment features or even audience requests (viewer's preferred brightness) can be adjusted. Various in S_im parameters are extracted by parameter extractor 950 and Some of these will be useful to derive a good LDR view.

Thus, in this system implementation, LDR is first used as the color mapping unit. Performing an additional color mapping via 953 and then, dynamic range conversion via brightness mapper 954 via HDR7 (not exactly original HDR, but a lot on the receiver side) close approximation RecýHDR) can be derived and this Outputs an LDR Video for the linkable LDR monitor 956. This situation such scenes, but even some scenes have some higher brightness The fact that its content is generated from some specular reflections is still a major shows improved quality, more natural and extraordinary visualization will give. If you want to trim or quantize on a display screen, for example If artifacts such as errors are seen as irritating, at least we Driving such display screens with the correct type of image We want to be proficient and thus allow the visualization display screen. Figure 10 shows which type, still following our EOTF technology. LDR_CONT/lm_l is another indication that the image must be actually encoded. It shows one specific example of a system built according to philosophy.

The dynamic range converter on the transmission side is 1001 and the Y”u”v” rec preset is 00 Similar components are similar to those in Figure 97. However, now we have on disk 28834.1284 We will use a Tm_1 with an LDR view. Thus, to obtain LDR Video In 1001, the log-gamma function was applied to different parameters rho and gamma. In the HDR view in the S_im scenario of Figure 99, by using to a higher equivalent gamma (our standard rho, gamma the equivalent gamma of our parameters, i.e. just a plane gamma function when used, L=V^gamma is about 7), but other scenes For , they may also be lower. In any case, now we typically We will only use reversible brightness mapping functions. and our EOTF and its reverse OETF portions fulfill this. LDR image (although typically still Y7u°v, rather than YCrCb) again with a regular The video goes to the encoder. Competent to perform any operation For legacy systems that do not have a YCrCb encoding, for example, if one You can view a single image in a particular application that encodes colors as Y'uv colors. If the person chooses to encode, the person still sees Y'uv as if he were a stranger. It behaves as if it were a colored YCrCb image. Thus, to achieve HDR From S_iml to LDR, a dynamic range is created to get the desired chronatic view. sequentially by upconverter 1050 and a color converter 1051 is processed and a display screen adjustment unit 1052 is reconstructed. “HDR” is a specific range starting from a reference range such as [0-5000]. Performs processing for the display screen. So we take a parameter We have a set of functions that can be controlled by . Rater For example, by turning a button, images can be displayed You can easily derive the optimal a-value for the shooting/operating situation and The automatic image analysis algorithm determines an Optimal a-value similarly can choose. And we also encode video in an image or to store the rater-selective a-value in the signal, such as a Hot or int (since we don't need many values, we use a-values as A*a+B and so on, different a-values, such as an 8-bit codeword This is achieved by defining a data type whose values can be assigned to We can easily code the relationship. Thus, for example, in a LUT-like signal, As an alternative to encoding the curve, applications of our technology are simply 28834.1284 They can encode the a-value (once or twice, to ensure security against data corruption). with the same value for or with different a-values for adaptability) and from then, if the used functions are pre-agreed at run time If they are not agreed upon in the standard, but are agreed upon within the standard, the purchasing party It will immediately know which real function is associated with its value.

Figure 11 shows an HDR image of the person, taking into account the teachings of the current application. shows an example of how it can encode the signal 1100. We, an HDR scene We assume that we code a set of dynamic range views on and For this, on the receiver side, we set a reference brightness range of 0.005-5000 nits. brightnesses to be visualized throughout their entirety are potentially We need a low master image with high dynamic range. Same At the same time, on the decoding side, we see that the same scene, the encoded HDR image 1101 and at least one to be determined on the basis of mapping functions. We would like to have the ability to redefine the low dynamic range image.

That being said, encoding/decoding blocks are functionalities in HEVCa. etc. will be Standard and thus, within this framework, HDR encoding We will focus on new colorimetric teachings to make this possible.

Thus, the 100 image pixel luina matrix of the image is a part of our main teaching. how luma codes relate to brightnesses in the reference brightness range definitions. For example, you are about to enter commands into the processor and find a loading to carry out any of its characteristic functions a series of steps (this translates into an intermediate language and a final processor language) (may include intermediate conversion steps such as ) In particular, computer program product, data on a carrier such as a recorder or tape, data present in a memory, data traveling through a network connection - wired or wireless - or can be implemented as program code on paper. From program code In addition, the characteristic data needed for the program can also be obtained from a computer It can be embedded as a program product. With computers, our data calculations 28834.1284 It should be clear that we mean that any device that can It could also be a mobile phone, for example. At the same time, device prompts may include computer-implemented versions of applications.

For some more advanced coding functions, a deviation of the log-gamma function Also, a deviation can be encoded in the number set 1107. For example, log-gamma an additive or multiplicative deformation along a portion of the function may contain a specification and a larger resp., a smaller gradient o is created in some sub-regions of the part and this is applied to various object regions of the image. resulting in more or less dedicated code. These numbers are also log-gamma can encode a functional transformation of our function, for example two parameter L1 and L2 at the set brightness or luma within a sub-range of the EOTF. and limits some parameters that define a transformation, for example ax^2+bx+c attracts and where x is a coordinate operating in the sub-range and a, b, C constants are various D3, They are encoded in number data fields D4,... What does the encoder function mean? will know that it has arrived because there are some pre-agreed rules for functional deformations. There will be residual mechanisms.

After that, other meta-data is HDR image 1100 encoded in the image signal. How to view an LDR looking image based on will define how it will be derived. This LDR image has all the features that can be achieved in HDR image. an image of less contrast showing the codes (with an additional gamma function mapped to LDR1) or a significant sub-range of the HDR scene and clips A contrast LDR that reserves most of the LDR luma codes available for There may be crooked clippings outside the view or 0 region.

Typically, to do a random mapping on lumas (for now (u, v) keeping the components the same), 1101 lumas of the HDR image and parametric Luma mapping is among the requirements in the LDR image to be co-coded. To contain a sufficiently precise LUT that encodes the l 1 10 format of the function It will be a meta-data information field. This function can have any format 28834.1284 and it doesn't even need to be monotonous (and no doubt it's also a brightness mapping, RGB max mapping or any brightness can be defined as correlation mapping). In addition, for example a color processing such as saturation mapping and finding per multiplicative saturation factor (from tone mapping from HDR image then luma dependent saturation modification for LDR image 1120 with a 1D LUT 1106 or more complex strategies that can be made, and this grader can detect some less bright objects in the LDR. or at least more colorful according to another saturation shifting philosophy gives permission to do so. Preferably, at the same time, we even have an LDR we encode the view, preferably the luma axis in the validation sub-ranges We still position the various captured scene objects along the When displayed directly, the LDR view content creator It will appear as preferred by. But we are also given by an additional optimized luma allocation function as a function We already code for specific monitoring application. In such a scenario, For example, for example, reference 5000 nits for raw input brightnesses above 5000 nits. equivalent display brightnesses for the final display screen within the brightness range. can rate for visualization.

The algorithmic components described in this text may not be used (fully or partially) in practice. as hardware (part of an application-specific IC) or as a dedicated digital signal as software running on the processor or a generic processor realizable.

From our presentation, the optional enhancements of the components are possible and feasibility in combination with other components and related devices how tools correspond to (optional) steps of methods It must be understandable by an expert. In this application, the word “device” means In its broadest sense, it nominally allows and enables the achievement of a specific purpose. So, for example, an IC (a small part of it) or a dedicated device (a 28834.1284 as a device with a display screen) or as part of a networked system is used. The word “regulation” also means, in its broadest sense, is intended to be used and thus, 0 alternately refers to a single device, a part of the device, a collection of cooperating devices (parts of them) may contain.

Commands are about to enter an operator and are characteristic of an invention. a series of loading steps to execute any of its functions. series (this is like translating between an intermediate language and a final processor language). may include conversion names) then to a generic or Special purpose processor any physical operation of a collection of instructions that enable the way the specifications should be understood to include the realization of A computer program product of existing applications. In particular, computer data on a carrier such as a program product or tape, stored in a memory existing data traveling through a network connection - wired or wireless can be implemented as data or program code on paper. Program Apart from the code, the characteristic data required for the program is also It may be embedded as a computer program product. With computers, our data It is clear that we mean that any device capable of performing calculations is possible. and it could also be, for example, a mobile phone. At the same time, Device claims may include computer-implemented versions of applications.

Some of the steps required for the operation of the method can be done in a computer program. Instead of defined ones such as data input and output steps in the product, the processor may already exist in its functionality.

The above-mentioned embodiments do not limit the invention but rather illustrate it. should be done. The scope of the invention is defined by the appended claims. Any reference marks between parentheses in the prompt are used to delimit the prompt. It is not intended. The word “containing” includes elements and elements not listed in a claim. It does not exclude the existence of issues. The "indefinite" word that comes before an element The word accusative does not exclude a plurality of such elements.

Claims (11)

CLAIMS l. A method of encoding a high dynamic range Video of images comprising the following steps: - inputting pixel colors of an input high dynamic range image, wherein the pixel colors have information of a luminance and a chromaticity; - applying an inverse of a mapping function to derive a luma code (v) of the luminance of a pixel color, with a first defined as P“1 in which p° is a tuning constant and V° is the luma code corresponding to a luminance to be encoded. Predetermined to include a partial function and a second partial partial mapping defined as L = LmPy in which Lm° is a peak luminance of a predefined reference display screen, i.e. a constant preferably equal to 2.4, - containing luma codes outputting a matrix of pixels having a color coding. 2. A method of encoding a high dynamic range video of images according to claim 1, wherein Lm is equal to 5000 nits. 3. A method of encoding a high dynamic range video of images according to Istein 1 or 2, wherein p is equal to 25. 28834.1284 . In Rec encoding y of the gamma function. A method of encoding a high dynamic range video of images according to claim 1 or 2, wherein the images are formed from an equivalent of 709° and a function y 2.4. . wherein the parameters p and y are further optimized to yield a coded image that appears good to a human color grader on a 100 nits display screen, and wherein at least one of the parameters p and y is preferentially optimized by a human color grader, A method of encoding a high dynamic range video of images according to one of the preceding claims. . A method of encoding a high dynamic range video of images according to one of the preceding claims, wherein the chromaticity coordinates (u,v) of the color coding are defined as referenced from a CIE XYZA representation of the colors of the pixels in the high dynamic range image (HDR_ORIG) by means of fractional equations of the following type: a... ] as constants and optionally . A method of encoding a high dynamic range video of images as claimed in any one of the preceding claims, wherein the chromaticity coordinates (u, V) are preferably defined relative to a predetermined white point such as D65. . A method of encoding a high dynamic range video of images according to one of the preceding claims, wherein an image signal (S_im) is formed comprising a pixel matrix image encoded with a color component that is the luma code and therein associated with metadata comprising at least one of the parameters p and y. method. 9. A video coding device for encoding a high dynamic range video of images, comprising: an input for obtaining pixel colors of the high dynamic range image, wherein the pixel colors have information of a luminance and a chromaticity; - a gradation management unit (202) arranged for applying an inverse of a mapping function to derive a luma code (v) of the luminance of a pixel color, in which p7 is a tuning constant and v is the luma code corresponding to a luminance to be coded, A first part function defined as P : (li-'11) 9 and a second part defined as L = LmPy where Lm" is a peak luminance of a predefined reference display screen, that is, a constant preferably equal to 2.4" configured to encode and transmit an image signal (S_im) comprising a pixel matrix image encoded with a color component being the luma code and associated therewith with metadata comprising at least one of the parameters p and y An encoder (210) connected to a video transmission link (221), which may be connected to a video memory or network. 10. A video coding device as claimed in claim 99, including a user interface unit (203) that allows a human rater to select a specific value of p and/or ya. 11. A video as claimed in claim 9, comprising an automatic image analysis unit (227) arranged to determine a certain value of p and/or y7 based on a statistical analysis of the brightness of objects present in at least one of the high dynamic range images (HDR_ORIG). coding device. A video coding device as claimed in any of the coding device claims, wherein the gradation management unit (202) is arranged to determine the chromatic components of the pixels of the high dynamic range image (HDR_ORIG) as follows: and ””“w a...l as constants and optionally , with the following values: A video decoding device (301) for decoding a high dynamic range video of encoding images (S_im), comprising: - receiving the high dynamic range image encoding (S_im) and resulting therefrom from an encoding method as claimed in claim an receiving and formatting unit (388) arranged to derive an image coding (lm_l) containing the luma codes to be processed; - a color mapping unit (305) arranged to implement a color mapping strategy to derive a high dynamic range image (REC_HDR) from the image coding lm_l, wherein the color mapping unit assigns pixel lumas v in the image coding (lm_l) to the image of the high dynamic range image (REC_HDR). ) to obtain the brightnesses of pixels It is arranged to implement a predetermined mapping function including a second portion partial mapping defined as L = LmPy where the display screen has a peak luminance and y”nm is preferably a constant equal to 2.4. The image coding S_im is arranged to implement a highly dynamic A video decoding device (301) as claimed in claim 133, for decoding intermittent video, wherein: the receiving and formatting unit (388) is used to derive at least one of the parameters rho, gamma or Lm from the high dynamic range image coding (S_im). is edited. In which the color mapping unit 305 performs a transformation to obtain the chromaticity coordinates (u, v) for the pixel colors and to map the Tm_1 pixel colors of the image together with the information of the luminance u and V components into a universal color representation such as a CIE XYZ color representation or RGB-like A video decoding device (301) for decoding a high dynamic range video of the image encoded S_im as claimed in one of the above image decoding device claims, further configured to apply to a device dependent color representation. In which the color mapping unit 305 is added to implement a second color mapping strategy using additional co-encoded color mapping parameters as meta-data that determines a color mapping to an image with a dynamic range difference from the dynamic range defined by the high dynamic range image (REC_HDR). A video decoding device (301) for decoding a high dynamic range video of the image coding (S_im) as claimed in one of the above video decoding device claims, embodied as . 17. A display screen comprising the video decoding device as claimed in any of the above decoding device claims. 18. A method of video decoding comprising: - taking an encoded high dynamic range video of images (S_im) and therefrom deriving an image encoding (lm_l) to be processed, and - deriving from the image encoding (Im_l) a high dynamic range image (REC_HDR) Performing color mapping through the application of a color mapping strategy such as , where p7 is an adjustment constant and Vin is used in the mapping function to obtain the brightnesses L of the pixels of the high dynamic range image (REC_HDR) on the pixel lumas v in the color mapping unit image coding (Im_l). is the luma code corresponding to a luminance to be encoded, a first part function defined as L : It is arranged to implement a predefined mapping function, including a second part partial mapping, defined as LmPy.