High dynamic range content distribution display method and video processing apparatus

Abstract

The application relates to the technical field of digital film content production, distribution and screening, and discloses a high dynamic range content distribution display method and a video processing device. A standard dynamic range master is used as the only main asset, physical baseline information is generated, candidate residual information is extracted, semantic gating decision is performed, auxiliary data containing the physical baseline information and selectively containing conditional gain information is output, and a high dynamic range signal is reconstructed according to the corresponding mode selected according to the auxiliary data, so that the single master is adapted to different display capability screening terminals, and the problems of high cost and poor consistency of the existing multi-version distribution mode are solved. Meanwhile, the semantic gating mechanism is used to eliminate the AI illusion risk, the adaptive auxiliary data structure is used to reduce the transmission bandwidth, and the distribution efficiency, screening safety and content verifiability are considered.

Description

Technical Field

[0001] This application relates to the field of integrated digital film content production, distribution and screening technology, specifically to a high dynamic range content distribution and display method and a video processing device. Background Technology

[0002] With the increasing prevalence of high dynamic range (HDR) display devices, the film industry faces the challenge of efficiently and securely distributing HDR content. Current technical solutions have the following three core issues: 1. The failure of the single master strategy and the generational mismatch dilemma.

[0003] In the early stages of digital cinema development, one of its core objectives was to establish a "single master" strategy. This strategy, through a unified digital cinema package format adapted to all cinemas, greatly reduced the physical copies and logistical costs associated with the film era. However, with the significant improvements in peak brightness and contrast of next-generation projection equipment such as LED direct-view displays, a serious generational mismatch has emerged between content and display terminals.

[0004] This technological gap has shattered the idealized single master version system. To adapt to display terminals with varying capabilities, the existing distribution system has been forced to degenerate into a multi-version distribution model. This model requires the separate production and distribution of completely independent digital cinema packages for SDR (Standard Dynamic Range) and HDR theaters. This not only multiplies the costs of production, storage, transmission, and key management, but also, due to the lack of a unified mapping standard, causes differences in image detail and brightness expression for the same film on different projection systems. This structurally weakens projection consistency, plunging the industry back into the chaos of multi-version management. The existing multi-version distribution model essentially compensates for differences in display capabilities by artificially replicating content assets, failing to address the mapping mechanism between content expression and physical display capabilities. Instead, it introduces new systemic risks such as unverifiable asset consistency and untraceable versions.

[0005] 2. Security risks arising from black-box enhancement.

[0006] While emerging AI-based end-to-end HDR conversion technologies can produce stunning effects, they carry risks of being "unexplainable" and "uncontrollable." Existing solutions often employ end-to-end pixel generation or style transfer methods, resulting in outputs that typically lack a decomposable and traceable physical relationship with the input signals. This makes it difficult to determine whether the enhancement stems from changes in display capabilities or from subjective algorithmic generation. Even with the introduction of dynamic metadata mechanisms, the underlying assumption remains that the enhancement behavior is predetermined and always effective, failing to provide a mechanism for determining "whether enhancement should occur," thus failing to constrain the AI ​​illusion problem at the system level.

[0007] 3. Data redundancy caused by fixed overhead.

[0008] Existing dynamic metadata schemes typically employ fixed-bandwidth transmission strategies, such as transmitting a full-resolution gain map per frame. However, film content exhibits significant non-uniformity; statistics show that over 90% of film runtime, including dialogue and natural scenery scenes, can be accurately represented by a simple global tone map using a physical baseline, without the need for complex local gains. Forcing the transmission of the full gain map data results in substantial storage and bandwidth waste.

[0009] In summary, there is an urgent need to propose an HDR content distribution and display solution that uses a single SDR master as the core, combining security and transmission efficiency, in order to break through the cost and consistency bottlenecks of existing multi-version distribution, solve the "black box" risk of AI enhancement, and reduce bandwidth consumption through an adaptive metadata mechanism to achieve seamless adaptation of different generations of projection terminals. Summary of the Invention

[0010] To address the aforementioned issues, this application provides a high dynamic range (HDR) content distribution and display method, an HDR content display method, an auxiliary data structure, and a video processing device. It only requires the distribution of a single standard SDR master to adapt to both SDR and HDR playback environments. Through a semantic gating mechanism, it eliminates AI illusions and blind enhancements, preventing irreversible local enhancements to the image without clear physical or semantic basis.

[0011] The technical solution adopted in this application is as follows: In a first aspect, this application provides a high dynamic range content distribution and display method, applied to a video processing apparatus comprising a production end, a transmission end, and a projection end connected in sequence, the method comprising: Step S1: The production end uses the standard dynamic range master as the main asset and generates physical baseline information for global monotonic mapping based on the standard dynamic range master. Step S2: The production end generates candidate residual information of the reference high dynamic range image relative to the high dynamic range baseline signal corresponding to the physical baseline information; Step S3: The production end performs semantic gating decision based on the candidate residual information; wherein, the decision is based on the energy statistics index of the candidate residual information and the structural correlation index with the standard dynamic range master image to determine whether to allow conditional gain information to enter the auxiliary data. Step S4: The production end generates auxiliary data based on the decision result; when the decision is allowed, auxiliary data containing physical baseline information, conditional gain information and status flag fields is generated; when the decision is disallowed, auxiliary data containing physical baseline information and status flag fields is generated. Step S5: The transmission end defines an adaptive auxiliary data structure and synchronously transmits the auxiliary data and the standard dynamic range master to the projection end. Step S6: The projection end reconstructs the high dynamic range signal based on the auxiliary data; wherein, when conditional gain information exists, a multiplicative gain is applied to the high dynamic range baseline signal in the linear optical domain in the linear optical domain; when conditional gain information does not exist, a high dynamic range signal equivalent to the high dynamic range baseline signal is output.

[0012] Secondly, this application also provides a video processing apparatus for performing the above-described high dynamic range content distribution and display method, comprising: The physical baseline generation module is used to generate physical baseline information for global monotonic mapping based on the standard dynamic range master as the main asset. The residual information generation module is used to generate candidate residual information of the reference high dynamic range image relative to the high dynamic range baseline signal corresponding to the physical baseline information; The semantic gating module is used to perform semantic gating decisions based on candidate residual information. The decision is based on the energy statistics index of the candidate residual information and the structural correlation index with the standard dynamic range master image to determine whether to allow conditional gain information to enter the auxiliary data. The auxiliary data generation module is used to generate auxiliary data based on the decision result; when the decision is to allow, it generates auxiliary data containing physical baseline information, conditional gain information, and status flag fields; when the decision is to disallow, it generates auxiliary data containing physical baseline information and status flag fields. The encapsulation and synchronization module defines an adaptive auxiliary data structure to synchronously transmit auxiliary data and the standard dynamic range master to the projection end. The reconstruction module is used to reconstruct the high dynamic range signal based on auxiliary data. When conditional gain information exists, a multiplicative gain is applied to the high dynamic range baseline signal in the linear optical domain. When conditional gain information does not exist, the output is a high dynamic range signal that is equivalent to the high dynamic range baseline signal.

[0013] Thirdly, this application also provides a high dynamic range content distribution method, applied to the production end of the aforementioned video processing device, the method comprising: Using the standard dynamic range master as the main asset, physical baseline information for global monotonic mapping is generated based on the standard dynamic range master; Generate candidate residual information of the reference high dynamic range image relative to the high dynamic range baseline signal corresponding to the physical baseline information; Semantic gating decisions are performed based on candidate residual information; wherein, the decision is based on the energy statistics index of the candidate residual information and the structural correlation index with the standard dynamic range master image to determine whether conditional gain information is allowed to enter the auxiliary data; Auxiliary data is generated based on the judgment result; when the judgment is allowed, auxiliary data containing physical baseline information, conditional gain information, and status flag fields is generated; when the judgment is disallowed, auxiliary data containing physical baseline information and status flag fields is generated.

[0014] Fourthly, this application also provides an auxiliary data structure applied to the transmission end of the aforementioned video processing apparatus, comprising: A parameterized lookup table encoded with physical baseline information representing a global monotonic mapping; When the status flag field indicates that conditional gain information is included, the auxiliary data structure includes a parameterized lookup table and conditional gain information; wherein, the conditional gain information is subjected to resolution scaling, includes region of interest block masking, quantization and compression coding; When the status flag field indicates that conditional gain information is not included, the auxiliary data structure only includes a parameterized lookup table; The auxiliary data structure is encapsulated in an MXF container and establishes a time axis synchronization relationship with the standard dynamic range master through KLV encoding.

[0015] Fifthly, this application also provides a high dynamic range content display method, applied to the projection end of the aforementioned video processing device, comprising: Receive standard dynamic range master and auxiliary data; When the status flag field indicates that conditional gain information is included, a multiplicative gain map is generated based on the conditional gain information. The nonlinear transformation of the standard dynamic range master is removed by inverse electro-optic conversion function to obtain a linear optical domain signal; A linear optical domain high dynamic range baseline signal is generated based on the linear optical domain signal and physical baseline information, and a pixel-by-pixel multiplicative operation is performed with the multiplicative gain map to obtain the linear optical domain high dynamic range signal; wherein, the multiplicative operation is a linear operator; The linear optical domain high dynamic range signal is converted into a high dynamic range signal by inverse electro-optic conversion function transformation. When the status flag field indicates that conditional gain information is not included, a high dynamic range baseline signal is generated based on the standard dynamic range master and physical baseline information. The high dynamic range signal is equivalent to the high dynamic range baseline signal.

[0016] The above-mentioned technical solution adopted in this application can achieve the following beneficial effects: Single-asset workflow: This fundamentally eliminates the dual-version distribution model. The SDR master copy serves as the sole primary asset, and with lightweight metadata, it delivers perfect performance on HDR screens. When playing on an SDR screen, the metadata is ignored, achieving perfect backward compatibility.

[0017] Absolute security: AI is effectively constrained through a physical baseline and semantic gating mechanism. HDR enhancement is transformed from a default behavior to a conditional behavior requiring proven justification, ensuring the entire projection chain meets the deterministic and traceability requirements of professional film projection. In ambiguous areas where enhancement is uncertain, it automatically reverts to the physical baseline, ensuring accurate and clean images even in the worst-case scenario.

[0018] Extremely high data transmission efficiency: Leveraging the spatiotemporal redundancy of film content, bandwidth is consumed only to transmit the gain map at moments truly requiring high dynamic range (such as highlights of less than 10%), while at other times only a very small LUT (parameterized lookup table) is transmitted. This on-demand allocation strategy greatly reduces the pressure on existing cinema server I / O.

[0019] High verifiability: Since the residual (i.e., gain map) and the baseline (i.e., LUT) are stored separately, third parties (such as directors or producers) can intuitively view and verify the gain map, clearly know which parts of the picture have been changed, and enhance the transparency of the technical solution. Attached Figure Description

[0020] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings: Figure 1 A flowchart illustrating a high dynamic range content distribution and display method according to an embodiment of this application is shown; Figure 2 A logical schematic diagram of semantic gating decision according to an embodiment of this application is shown; Figure 3 A schematic diagram of the structure of an auxiliary data structure according to an embodiment of this application is shown; Figure 4 A schematic diagram of the structure of a video processing apparatus according to an embodiment of this application is shown; Figure 5 A flowchart illustrating a high dynamic range content distribution method according to an embodiment of this application is shown; Figure 6 A flowchart illustrating a high dynamic range content display method according to an embodiment of this application is shown. Detailed Implementation

[0021] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of this application will be clearly and completely described below in conjunction with specific embodiments and corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0022] To clarify the scope of protection of this application, the following core terms are defined.

[0023] Standard Dynamic Range Master: A standard dynamic range video asset that conforms to digital cinema standards (such as the DCI specification) and is the sole primary distribution asset of this application.

[0024] Physical baseline information: A data-driven carrier describing the global monotonic mapping relationship from the standard dynamic range master to the high dynamic range baseline signal, representing the inevitable physical result of the brightness expansion of the display device.

[0025] Candidate residual information: The local gain or difference components between the reference high dynamic range image and the high dynamic range baseline signal corresponding to the physical baseline information, which are represented as residual maps.

[0026] Semantic gating decision: A binary control mechanism that judges candidate residual information and decides whether to allow it to enter the distribution / rendering chain.

[0027] Rollback mode: Auxiliary data includes physical baseline information.

[0028] Activation mode: Auxiliary data includes physical baseline information and conditional gain information.

[0029] Display degrees of freedom: Local brightness variations that exceed the range of global monotonic mapping but can be supported by semantic or physical features (also known as conditional degrees of freedom).

[0030] Linear optical domain map: RGB image data that conforms to optical linearity obtained after performing an inverse electro-optical conversion function (EOTF) transformation on a standard dynamic range master. It is the basic processing carrier for eliminating nonlinear color distortion.

[0031] Linear optical domain high dynamic range baseline signal: The pixel data in the linear optical domain map is transformed according to the global monotonic mapping relationship corresponding to the physical baseline information to obtain a high dynamic range baseline signal in the linear optical domain. This signal does not contain any local gain correction.

[0032] Linear optical domain high dynamic range signal: The signal obtained by performing pixel-by-pixel multiplication operation between the linear optical domain high dynamic range baseline signal and the multiplicative gain map.

[0033] High Dynamic Range Baseline Signal: The baseline HDR signal obtained by converting the standard dynamic range master based on the physical baseline information.

[0034] High dynamic range signal: refers to the final HDR electrical signal output by the projection end.

[0035] Hereinafter, standard dynamic range will be represented by SDR, and high dynamic range by HDR.

[0036] The core technical solution adopted in this application is a single master release display method based on the theory of "constrained restoration." This method advocates for the controlled release of HDR potential by establishing an auditable physical baseline and semantic constraint mechanism, while respecting the original SDR narrative structure. It includes data generation methods on the production end, data structure definitions on the transmission end, and restoration display methods on the projection end.

[0037] Figure 1 A schematic flowchart of a high dynamic range content distribution and display method according to an embodiment of this application is shown. (Refer to...) Figure 1 As shown, the high dynamic range content distribution and display method proposed in this embodiment is applied to a video processing device comprising a production end, a transmission end, and a projection end connected in sequence. The method includes: Step S1: The production end uses the standard dynamic range master as the main asset and generates physical baseline information for global monotonic mapping based on the standard dynamic range master. Step S2: The production end generates candidate residual information of the reference high dynamic range image relative to the high dynamic range baseline signal corresponding to the physical baseline information; Step S3: The production end performs semantic gating decision based on the candidate residual information; wherein, the decision is based on the energy statistics index of the candidate residual information and the structural correlation index with the standard dynamic range master image to determine whether to allow conditional gain information to enter the auxiliary data. Step S4: The production end generates auxiliary data based on the decision result; when the decision is allowed, auxiliary data containing physical baseline information, conditional gain information and status flag fields is generated; when the decision is disallowed, auxiliary data containing physical baseline information and status flag fields is generated. Step S5: The transmission end defines an adaptive auxiliary data structure and synchronously transmits the auxiliary data and the standard dynamic range master to the projection end. Step S6: The projection end reconstructs the high dynamic range signal based on the auxiliary data; wherein, when conditional gain information exists, a multiplicative gain is applied to the high dynamic range baseline signal in the linear optical domain in the linear optical domain; when conditional gain information does not exist, a high dynamic range signal equivalent to the high dynamic range baseline signal is output.

[0038] I. Production side.

[0039] A core task in the production process is to generate physical baseline information and conditional gain information after semantic gating. This step essentially decomposes the HDR image into a superposition of a global mapping baseline and sparse residuals. This is expressed by the following formula (1): , formula (1); in, Indicates SDR master. This represents a globally monotonic mapping function. This represents the conditional gain information after semantic gating. This indicates an HDR signal.

[0040] To balance the efficiency of automation with the security required for professional applications, this application introduces dual constraints of physical baseline information and semantic gating decisions during the data generation stage at the manufacturing end.

[0041] 1. Physical baseline information generation.

[0042] A global monotonic mapping function is established from the SDR image to the reference HDR image. The physical baseline information generated based on this global monotonic mapping function represents the best HDR effect achievable only through global brightness stretching (without changing local contrast). Preferably, a regression or fitting method satisfying monotonic constraints is used to calculate the global monotonic mapping function. More preferably, an ordinal regression algorithm (PAVA) is used to calculate the global monotonic mapping function to maximize the interpretability of the baseline for the light energy distribution of the HDR image, thereby ensuring the rationality of the physical baseline information.

[0043] Specifically, input a frame of SDR image and the corresponding reference HDR image It should be noted that the reference HDR image is... The results can be derived from colorists' manual output, historical HDR version aligned frames, or high-precision AI model predictions. They are only used for production-end calculations and are not used as final release assets. High-precision AI model predictions are preferred.

[0044] Will and Flattened into a one-dimensional pixel array, the global monotonic mapping function is solved using the order-preserving regression algorithm (PAVA). The goal is to minimize the weighted squared error. This is expressed by the following formula (2): , formula (2); in, The index of the pixel represents the error calculated pixel by pixel; Indicates the first The weighting coefficients for each pixel are used to adjust the error weights for different regions; Indicates the first The pixel values ​​of an SDR image. Indicates the first The reference HDR image pixel values; This represents the set of all globally monotonic mapping functions.

[0045] Global monotonic mapping function Able to interpret reference HDR images More than 90% of the energy is distributed in the middle.

[0046] 2. Residual extraction.

[0047] Calculate reference HDR image The residual map R (candidate residual information) is obtained by comparing the residual between the HDR baseline signal and the physical baseline information. It is expressed by the following formula (3): , formula (3).

[0048] The residual map represents the local details that the deep neural network believes need to be added beyond the global stretch (such as glare from highlights or neon glow).

[0049] 3. Semantic gating decision.

[0050] Semantic gating is the core control logic of this application. Semantic gating is not used to select the enhancement method, but rather to determine whether any local enhancement behavior is allowed. When the decision condition is not met, any enhancement information that is not globally mapped is prohibited from being output. In other words, semantic gating determines whether candidate residual information has a physical or semantic basis, allowing only valid candidate residual information to pass, thus preventing AI illusions.

[0051] Figure 2 A logical schematic diagram of semantic gating decision according to an embodiment of this application is shown. (Refer to...) Figure 2 As shown, the specific implementation process of semantic gating decision is as follows.

[0052] 3.1 Energy verification.

[0053] Calculate the absolute value of the residual plot R. The high quantile energy statistical characteristics are taken as the energy statistical index. Preferably, the 95th or 99th percentile is used.

[0054] If the energy statistics meet the first preset condition, the decision is "allowed," and the system enters activation mode. If the energy statistics do not meet the first preset condition, the system enters a pending comprehensive scoring and verification phase.

[0055] For example, if Furthermore, if the spatial location is concentrated and corresponds to a strong light source in the image (such as laser, explosion, direct light source, etc.), then it must enter the active mode, otherwise the dynamic range will be lost.

[0056] 3.2 Calculation of structural correlation.

[0057] For residual maps R and SDR images High-pass filtering is performed on both, and the correlation metric of their gradient maps is calculated as a structural correlation index. Preferably, the Spearman rank correlation coefficient is used as the correlation measure.

[0058] If the structural relevance index meets the second preset condition, the decision is "allowed," and the system enters activation mode. If the structural relevance index does not meet the second preset condition, the system enters a pending comprehensive scoring verification phase.

[0059] For example, if Furthermore, with moderate energy, it can enter activation mode to enhance the realism of materials such as metallic reflections and glass textures.

[0060] 3.3 Comprehensive score verification.

[0061] The comprehensive score is determined based on energy statistics and structural correlation indicators. This can be expressed by the following formula (4): , formula (4); in, These are the lighting / texture weighting coefficients determined statistically based on the StEM2 dataset.

[0062] If the overall score meets the third preset condition, the decision is "allowed," and the system enters activation mode. If the overall score does not meet the third preset condition, the decision is "disallowed," and the system enters rollback mode.

[0063] For example, if If the frame contains valid semantics, it is determined to enter active mode. Otherwise, for the 90% of the ordinary diffuse area in the image, local enhancement is intercepted and only global mapping is applied, i.e., it enters fallback mode.

[0064] Through the aforementioned semantic gating decision, any candidate residual information is either explicitly classified as an interpretable display degree of freedom or considered an invalid signal without physical or semantic basis and eliminated, thus forming a binary decision closed loop for enhancement or not. This step not only filters out noise but also physically ensures that the image is only intervened when it has a clear semantic direction, eliminating the risk of AI illusion.

[0065] 4. Auxiliary data generation.

[0066] Based on the semantic gating decision, the production end generates auxiliary data and status flag fields for two modes (activation mode and rollback mode).

[0067] 4.1 Activation Mode Auxiliary Data.

[0068] When the decision is permissible, the auxiliary data includes: physical baseline information and conditional gain information, where the conditional gain information is a multiplicative gain map or a difference map.

[0069] When the decision is allowed, the production end simultaneously generates a status flag field corresponding to the decision, which is used to indicate that it contains conditional gain information.

[0070] The lowest bit (flag bit) of the status flag field, Is_Fallback=0, indicates the presence of conditional gain information.

[0071] The auxiliary data also includes validation fields independent of the conditional gain information. These validation fields include: gate mask digest, energy statistics, structural correlation metrics, and fallback reason codes, supporting rapid compliance checks. The validation fields can be parsed without decoding the conditional gain information to support rapid compliance checks and fallback reason tracing.

[0072] 4.2. Rollback Mode Auxiliary Data.

[0073] When the decision is not permitted, the auxiliary data only includes physical baseline information.

[0074] When the judgment result is not allowed, the production end simultaneously generates a status flag field corresponding to the judgment result to indicate that conditional gain information is not included.

[0075] The least significant bit of the status flag field, Is_Fallback=1, indicates that there is no conditional gain information.

[0076] The auxiliary data may also include validation fields. Validation fields include: gate mask digest and fallback reason code.

[0077] II. Transmission end.

[0078] The core task of the transmission end is to define an adaptive auxiliary data structure and distribute the auxiliary data synchronously with the SDR master to ensure efficient data transmission and compatibility with existing cinema systems.

[0079] This application defines a dynamic, hierarchical metadata signal structure to carry the above processing results. Adaptability is reflected in the "variability of structural state and carried content," that is, by encoding "whether there are allowed display degrees of freedom" as structural state, the data size becomes a function of the semantic complexity of the content; at the same time, for the conditional gain information segments that need to be transmitted, sparse representation and video coding can be used for compression.

[0080] 1. Definition of auxiliary data structures.

[0081] Figure 3 A schematic diagram of an auxiliary data structure according to an embodiment of this application is shown. (Refer to...) Figure 3 As shown, the globally monotonic mapping function obtained by solving... The physical baseline information is encoded as a 1D parametric lookup table (LUT), which contains 1024 floating-point numbers and has a data size of approximately 4KB.

[0082] Fallback mode structure: When all residuals in the current frame are intercepted by semantic gating decisions (i.e., the decision result is not allowed), the core data segment of the data packet only contains LUTs. The data volume in this mode is extremely small (e.g., less than 5KB / frame).

[0083] Activation mode structure: When valid semantic residuals exist, the core data segment of the data packet contains a LUT and quantized conditional gain information. In this mode, the amount of data dynamically changes with the complexity of the content, providing ultimate local dynamic range control.

[0084] To meet the stringent bandwidth and decoding efficiency requirements of theatrical releases, the conditional gain information preferably employs a "sparse compression representation based on video coding." This data segment contains the following key descriptive fields: GainMap_Resolution: In Embodiment 1 below of this application, Resolution_Scale is set to 1 / 4 (i.e., 1 / 16 of the pixel amount). This parameter is chosen based on the characteristic that the human visual system (HVS) may be less sensitive to brightness details than to texture edges over a wide dynamic range.

[0085] Tile_Map: Indicates the block mask of the effective gain region, encoding only the active region.

[0086] Codec_Profile: Indicates the video encoding configuration used, such as HEVC Low-delay P with a quantization depth of 10-bit in Example 1 below.

[0087] The specific processing flow for conditional gain information is as follows.

[0088] Resolution scaling: Scale the residual map R to 1 / 2, 1 / 4 or 1 / 8 of the SDR master; Includes region of interest block mask: Generates a Tile_Map to indicate the block mask of the effective gain region, encoding only the active region; Quantization: Log2 or PQ-delta quantization is used, and the quantization depth can be 10-16 bits; Compression encoding: HEVC series compression format (such as HEVC Intra or Low-delay P, etc.) is used.

[0089] Specifically, the following are three preferred embodiments of conditional gain information.

[0090] Example 1, Preferred Example: Key parameters: 1 / 4 resolution downsampling; quantization: 10-bit Log2 / PQ; encoding: HEVC Main10.

[0091] Technical effect: While maintaining a high fidelity of PSNR (peak signal-to-noise ratio) >50dB, the bandwidth is compressed to less than 0.12MB / frame, making it possible to achieve real-time decoding on existing DCI cinema server (IMB) hardware.

[0092] Example 2, Conservative - Very Low Risk: Gain map resolution: 1 / 8 of the main image resolution; quantization: 10-bit Log2; encoding: HEVC Intra.

[0093] Bitrate budget: 0.2~1.5MB / s on average.

[0094] Applicable scenarios: Scenarios requiring compatibility with older versions of IMB or extreme stability.

[0095] Example 3, Radical - High-end Customization: Gain map resolution: 1 / 2 main image resolution; quantization: 12-16 bit PQ-delta; encoding: HEVC Main10 / RExt.

[0096] Bitrate budget: 1.0~6.0 MB / s on average.

[0097] Applicable scenarios: Dolby Cinema or LED-exclusive high-specification release.

[0098] Thus, the specific format of the auxiliary data structure is shown in Table 1.

[0099] Table 1, Auxiliary Data Structure:

[0100] In addition, by integrating a bandwidth prediction module, automatic degradation can be performed according to the following priorities when the estimated gain data exceeds the preset transmission budget (such as in complex fireworks scenarios) to ensure the determinism of the output: Priority 1, reduce resolution (1 / 2→1 / 4→1 / 8). Priority 2, reduce quantization precision; Priority 3: Tighten the semantic gating threshold and reduce the number of activated tiles; Priority 4 triggers fragment rollback; the current GOP (Group of Pictures) only sends rollback mode auxiliary data.

[0101] 2. Encapsulation and synchronization mechanisms.

[0102] MXF container encapsulation: Auxiliary data is encapsulated into MXF files using the SMPTE 377M specification, serving as a private data stream or independent auxiliary data asset, and together with the J2K encoded data (lossless compression) of the SDR master, they form a Digital Cinema Package (DCP).

[0103] KLV encoding synchronization: KLV encoding is used to achieve hard synchronization between auxiliary data and SDR master. Each frame of video corresponds to a KLV data packet, and the timestamp of the data packet is strictly aligned with the frame timestamp of the SDR master to ensure frame-level synchronization at the playback end.

[0104] CPL Reference Registration: Auxiliary data assets are registered in the DCP Composite Playlist (CPL) to establish timecode alignment references with the main asset (SDR master), supporting automatic recognition and loading by the cinema server.

[0105] This design ensures that auxiliary data exists as an optional enhancement layer. In HDR playback systems that support loading auxiliary data assets, these assets are loaded to activate the enhancements; while in older systems that do not support this extension, the assets can be ignored or not loaded, while the main SDR assets can still be decoded and played normally, thus guaranteeing the physical compatibility of a single master copy across different generations of systems.

[0106] III. Projection Terminal.

[0107] The core task of the projection end is to analyze auxiliary data while maintaining color accuracy. Figure 1 Under the premise of consistency, reconstruct the HDR signal.

[0108] 1. Data parsing and synchronization.

[0109] The decoder at the projection end reads both the SDR master (J2K encoded) and the auxiliary data stream (KLV data packets encapsulated in MXF).

[0110] Parse the Is_Fallback flag in the KLV packet: When parsing a KLV packet, determine whether the current frame is in fallback mode or active mode based on the Is_Fallback flag in the packet header.

[0111] 2. Activation Mode.

[0112] The hardware decoder is invoked to dequantize the multiplicative gain map. When the multiplicative gain map is at a downsampled resolution (e.g., 1 / 4), the projection end can use higher-order interpolation (e.g., bi / cubic interpolation) or edge-preserving upsampling based on the gradient / structure information of the SDR master picture to reconstruct the full-resolution gain map.

[0113] The full-resolution gain map is multiplicatively superimposed onto the HDR baseline signal. The HDR reconstruction process is strictly performed in the linear optical domain to ensure basic hue consistency. While the basic dynamic range mapping is performed in the luminance domain, the "constrained restoration" strategy (linear optical domain operation) for the color dimension strictly defines that color dimension processing must be performed in the gamma-free linear optical domain to ensure the photometric criteria are met. That is, hue shift is strictly limited during the SDR to HDR conversion process.

[0114] Assume the input signal (SDR master) is in a standard cinema color gamut (such as DCI-P3 or Rec.2020) and has been converted into a linear light signal via EOTF (such as Gamma 2.6). The subsequently calculated luminance gain coefficients are then simultaneously applied to the three linear RGB channels. Unlike direct operation in the Gamma domain, maintaining the proportions in the linear domain physically ensures that the R:G:B vector direction remains unchanged, thereby eliminating hue / saturation distortion introduced by non-linear curves.

[0115] The specific steps are as follows: The first transformation sub-step is SDR linearization.

[0116] The nonlinear transformation of the SDR master is removed by inverse electro-optic conversion function to obtain a linear optical domain signal. This is expressed by the following formula (5): , formula (5); in, This represents the inverse transform of the Gamma 2.6 electro-optic conversion function. This represents a linear optical domain signal.

[0117] Multiplicative sub-step, brightness enhancement.

[0118] Generate linear optical domain HDR baseline signal based on linear optical domain signal and physical baseline information. This can be expressed by the following formula (6): , formula (6); Among them, because the transmission end will use the global monotonic mapping function The physical baseline information is encoded as a 1D parametric lookup table (LUT). Therefore, this formula represents the physical baseline information using a LUT. That is, the LUT is the engineering implementation of the physical baseline information, and f is the mathematical essence of the physical baseline information.

[0119] Linear optical domain HDR baseline signal A pixel-wise multiplicative operation is performed with the conditional gain information (full-resolution gain map) to obtain the linear optical domain HDR signal. This is expressed by the following formula (7): , formula (7); in, Indicates pixel-by-pixel multiplication. This represents the full-resolution gain graph.

[0120] The second conversion sub-step is the target display driver.

[0121] The linear optical domain HDR signal is converted into an HDR signal using the inverse electro-optical transformation function. This can be expressed by the following formula (8): , formula (8); in, This represents the inverse transform of the PQ curve.

[0122] 3. Rollback mode.

[0123] In rollback mode, the HDR baseline signal is directly generated based on the SDR master and physical baseline information, and the HDR signal is equivalent to the HDR baseline signal.

[0124] When the projection system does not implement or enable auxiliary data parsing, the SDR master can be directly decoded and displayed, and a display result consistent with the original master's intent can be obtained without relying on any external parameters. Therefore, the compatibility of this application stems from the integrity of the main asset itself, rather than relying on a specific playback strategy.

[0125] 4. Volume release allows color saturation to naturally increase as the dynamic range of brightness expands.

[0126] Boundary conditions: This release is primarily triggered in the highlight region above the diffuse white point (approximately 203 cd / m²) and is mapped to the HDR peak brightness.

[0127] Physical meaning: By utilizing the larger color volume container of HDR, the color potential suppressed by the SDR container is released, achieving physical-level compensatory restoration.

[0128] Cascade constraints of semantic gating: Volume release in the color dimension does not occur unconditionally; it must be subject to the cascade constraints of the aforementioned semantic gating decisions. That is, only when the luminance residual is determined to be a valid semantic (flag = activated) is the corresponding region allowed to break through the SDR color gamut limit for volume release; otherwise, color saturation should be strictly limited within the baseline range to prevent noise from being disguised as high color capacity information.

[0129] To verify the engineering feasibility and image quality restoration capability of the above configuration (especially Example 1), actual tests were conducted on the StEM2 test sequence. The results show that when using a 1 / 4 resolution downsampling + HEVC compression strategy: Bandwidth reduction: The data volume is reduced from 14.0 MB / frame in the original Float16 to 0.04~0.12 MB / frame (average compression ratio greater than 100 times), which fully meets the engineering budget of typical cinema server I / O and playback link (equivalent to 1.0~3.0 MB / s@24fps).

[0130] Image quality fidelity: Within a specified evaluation domain, the error between the output and Ground Truth HDR meets a preset threshold; for example, the PSNR reaches the order of 50dB on Active frames (such as Frame 05947), and the P99.9 error of the luminance histogram distribution is negligible (e.g., less than 1%). This confirms that this application, through a combination of "semantic gating + spatial downsampling" strategy, can achieve industrial-grade HDR reproduction accuracy at extremely low bandwidth cost.

[0131] In addition, the projection end can quickly verify the compliance of the enhancement effect through the verification fields of the auxiliary data: compare the gate mask digest with the local calculation results to confirm the consistency of the gain region; check the energy statistics and structural correlation indicators to confirm that the enhancement behavior conforms to the physical logic; if the verification fails, it will automatically switch to the fallback mode.

[0132] Figure 4 A schematic diagram of a video processing apparatus according to an embodiment of this application is shown. (Refer to...) Figure 4 As shown, based on the same concept, the video processing apparatus A for performing the high dynamic range content distribution and display method proposed in this embodiment includes: The physical baseline generation module A1 is used to generate physical baseline information with a global monotonic mapping based on the standard dynamic range master as the main asset. The residual information generation module A2 is used to generate candidate residual information of the reference high dynamic range image relative to the high dynamic range baseline signal corresponding to the physical baseline information; The semantic gating module A3 is used to perform semantic gating decisions based on candidate residual information. The decision is based on the energy statistics index of the candidate residual information and the structural correlation index with the standard dynamic range master image to determine whether to allow conditional gain information to enter the auxiliary data. The auxiliary data generation module A4 is used to generate auxiliary data based on the decision result; when the decision is allowed, auxiliary data containing physical baseline information, conditional gain information and status flag fields is generated; when the decision is disallowed, auxiliary data containing physical baseline information and status flag fields is generated. The A5 encapsulation and synchronization module defines an adaptive auxiliary data structure to synchronously transmit auxiliary data and the standard dynamic range master to the projection end. The reconstruction module A6 is used to reconstruct the high dynamic range signal based on auxiliary data; wherein, when conditional gain information exists, a multiplicative gain is applied to the high dynamic range baseline signal in the linear optical domain in the linear optical domain; when conditional gain information does not exist, a high dynamic range signal equivalent to the high dynamic range baseline signal is output.

[0133] In some alternative implementations, the physical baseline information is generated by a regression or fitting method that satisfies monotonic constraints, by solving a global monotonic mapping function from the standard dynamic range master to the reference high dynamic range signal by minimizing the weighted squared error.

[0134] In some optional implementations, the energy statistics index is: the high quantile energy statistics of the candidate residual information; the structural correlation index is: the correlation measure between the candidate residual information and the gradient map of the standard dynamic range master image.

[0135] In some optional implementations, the conditional gain information is a multiplicative gain map; wherein the conditional gain information is subjected to resolution scaling, including region of interest block masking, quantization and compression coding, the resolution scaling ratio is 1 / 2, 1 / 4 or 1 / 8, the quantization method is log2 or PQ-delta, and the compression coding format is HEVC series.

[0136] In some optional implementations, when the decision is allowed, the auxiliary data also includes independent verification fields; wherein the verification fields include: gating mask digest, energy statistics, structural correlation index, and backoff reason code.

[0137] In some alternative implementations, the auxiliary data is organized by frame, image group, or shot segment.

[0138] In some alternative implementations, if the multiplicative gain map corresponding to the conditional gain information is at downsampling resolution during reconstruction, then higher-order interpolation or edge-preserving upsampling based on the standard dynamic range master master gradient or structural information is used to reconstruct the full-resolution gain map.

[0139] In some optional implementations, the reconstruction with conditional gain information includes: a first conversion sub-step, removing the nonlinear transformation of the standard dynamic range master through an inverse electro-optic conversion function to obtain a linear optical domain signal; a multiplicative operation sub-step, generating a linear optical domain high dynamic range baseline signal based on the linear optical domain signal and physical baseline information, and performing a pixel-by-pixel multiplicative operation with the conditional gain information to obtain a linear optical domain high dynamic range signal; and a second conversion sub-step, converting the linear optical domain high dynamic range signal into a high dynamic range signal through an inverse electro-optic conversion function.

[0140] In some alternative implementations, the reconstruction process is configured to release the color saturation of the highlight regions using the extended dynamic range while maintaining hue consistency with the standard dynamic range master image; wherein the trigger threshold for color volume release is anchored to the diffuse white point, and the magnitude of color enhancement is constrained by the cascading of semantic gating results.

[0141] Figure 5 A flowchart illustrating a high dynamic range content distribution method according to an embodiment of this application is shown. (Refer to...) Figure 5 As shown, based on the same concept, the high dynamic range content distribution method proposed in this embodiment is applied to the production end of video processing device A, and the method includes: Step S7: Using the standard dynamic range master as the main asset, generate physical baseline information for global monotonic mapping based on the standard dynamic range master; Step S8: Generate candidate residual information of the reference high dynamic range image relative to the high dynamic range baseline signal corresponding to the physical baseline information; Step S9: Perform semantic gating decision based on candidate residual information; wherein, the decision is based on the energy statistics index of candidate residual information and the structural correlation index with the standard dynamic range master image to determine whether conditional gain information is allowed to enter the auxiliary data. Step S10: Generate auxiliary data based on the decision result; if the decision is allowed, generate auxiliary data containing physical baseline information, conditional gain information, and status flag fields; if the decision is disallowed, generate auxiliary data containing physical baseline information and status flag fields.

[0142] Based on the same concept, this application proposes an auxiliary data structure for use at the transmission end of the aforementioned video processing device A, comprising: Physical baseline information characterizing the global monotonic mapping; A status flag field indicating the presence of conditional gain information; wherein, when the status flag field indicates that conditional gain information is included, the conditional gain information is resolution-scaled, includes region of interest block mask, quantization and compression coding; when the status flag field indicates that conditional gain information is not included, the auxiliary data only includes physical baseline information. The auxiliary data structure is encapsulated in an MXF container and establishes a time axis synchronization relationship with the standard dynamic range master through KLV encoding.

[0143] Figure 6 A flowchart illustrating a high dynamic range content display method according to an embodiment of this application is shown. (Refer to...) Figure 6 As shown, based on the same concept, the high dynamic range content display method proposed in this embodiment is applied to the projection end of video processing device A, and the method includes: Step S11: Receive the standard dynamic range master and auxiliary data; Step S12: When the status flag field indicates that conditional gain information is included, generate a multiplicative gain map based on the conditional gain information. Step S13: Remove the nonlinear transformation of the standard dynamic range master through inverse electro-optic conversion function to obtain a linear optical domain signal; Step S14: Generate a linear optical domain high dynamic range baseline signal based on the linear optical domain signal and physical baseline information, and perform a pixel-by-pixel multiplicative operation with the multiplicative gain map to obtain the linear optical domain high dynamic range signal; wherein, the multiplicative operation is a linear operator; Step S15: Convert the linear optical domain high dynamic range signal into a high dynamic range signal by using the inverse electro-optic conversion function. Step S16: When the status flag field indicates that it does not contain conditional gain information, a high dynamic range baseline signal is generated based on the standard dynamic range master and physical baseline information. The high dynamic range signal is equivalent to the high dynamic range baseline signal.

[0144] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application, and should all be included within the protection scope of this application.

Claims

1. A method for distributing and displaying high dynamic range content, characterized in that, Applied to a video processing apparatus comprising a production end, a transmission end, and a projection end connected in sequence, the method includes: Step S1: The production end uses the standard dynamic range master as the main asset and generates physical baseline information for global monotonic mapping based on the standard dynamic range master. Step S2: The production end generates candidate residual information of the reference high dynamic range image relative to the high dynamic range baseline signal corresponding to the physical baseline information; Step S3: The production end performs semantic gating decision based on the candidate residual information; wherein, the decision is based on the energy statistics index of the candidate residual information and the structural correlation index with the standard dynamic range master image to determine whether to allow conditional gain information to enter the auxiliary data. Step S4: The production end generates auxiliary data based on the decision result; when the decision is allowed, auxiliary data containing physical baseline information, conditional gain information and status flag fields is generated; when the decision is disallowed, auxiliary data containing physical baseline information and status flag fields is generated. Step S5: The transmission end defines an adaptive auxiliary data structure and synchronously transmits the auxiliary data and the standard dynamic range master to the projection end. Step S6: The projection end reconstructs the high dynamic range signal based on the auxiliary data; wherein, when conditional gain information exists, a multiplicative gain is applied to the high dynamic range baseline signal in the linear optical domain in the linear optical domain; when conditional gain information does not exist, a high dynamic range signal equivalent to the high dynamic range baseline signal is output.

2. The high dynamic range content distribution and display method according to claim 1, characterized in that, The physical baseline information in step S1 is generated by a regression or fitting method that satisfies monotonic constraints, and the global monotonic mapping function from the standard dynamic range master to the reference high dynamic range signal is solved by minimizing the weighted squared error.

3. The high dynamic range content distribution and display method according to claim 1, characterized in that, The energy statistics index in step S3 is: the high quantile energy statistics characteristics of the candidate residual information; The structural correlation index in step S3 is a measure of the correlation between candidate residual information and the gradient map of the standard dynamic range master image. The decision in step S3 includes: when the energy statistics index meets the first preset condition, or the structural correlation index meets the second preset condition, or the comprehensive score determined based on the energy statistics index and the structural correlation index meets the third preset condition, the decision is to allow; Otherwise, it is not allowed.

4. The high dynamic range content distribution and display method according to claim 1, characterized in that, The conditional gain information in step S4 is a multiplicative gain map; wherein, the conditional gain information is subjected to resolution scaling, including region of interest block masking, quantization and compression coding, the resolution scaling ratio is 1 / 2, 1 / 4 or 1 / 8, the quantization method is log2 or PQ-delta, and the compression coding format is HEVC series.

5. The high dynamic range content distribution and display method according to claim 1, characterized in that, When the decision in step S4 is to allow, the auxiliary data also includes independent verification fields; wherein, the verification fields include: gate mask digest, energy statistics, structural correlation index and backoff reason code.

6. The high dynamic range content distribution and display method according to claim 1, characterized in that, The auxiliary data in step S5 is organized by frame, image group or shot segment.

7. The high dynamic range content distribution and display method according to claim 1, characterized in that, During step S6 reconstruction, if the multiplicative gain map corresponding to the conditional gain information is at downsampling resolution, then higher-order interpolation or edge-preserving upsampling based on the standard dynamic range master drawing gradient or structural information is used to reconstruct the full-resolution gain map.

8. The high dynamic range content distribution and display method according to claim 1, characterized in that, Reconstruction in step S6 when conditional gain information exists includes: The first conversion sub-step removes the nonlinear conversion of the standard dynamic range master through the inverse electro-optic conversion function to obtain a linear optical domain signal; The multiplicative operation sub-step generates a linear optical domain high dynamic range baseline signal based on the linear optical domain signal and physical baseline information, and performs a pixel-by-pixel multiplicative operation with the conditional gain information to obtain the linear optical domain high dynamic range signal. The second conversion sub-step transforms the linear optical domain high dynamic range signal into a high dynamic range signal through an inverse electro-optic conversion function.

9. The high dynamic range content distribution and display method according to claim 1, characterized in that, The reconstruction process is configured to release the color saturation of the highlight region by utilizing the extended dynamic range while maintaining the hue consistency of the standard dynamic range master image; wherein the trigger threshold for color volume release is anchored to the diffuse white point, and the color enhancement magnitude is constrained by the cascade of semantic gating results.

10. A video processing apparatus for implementing the high dynamic range content distribution and display method according to any one of claims 1 to 9, characterized in that, include: The physical baseline generation module is used to generate physical baseline information for global monotonic mapping based on the standard dynamic range master as the main asset. The residual information generation module is used to generate candidate residual information of the reference high dynamic range image relative to the high dynamic range baseline signal corresponding to the physical baseline information; The semantic gating module is used to perform semantic gating decisions based on candidate residual information. The decision is based on the energy statistics index of the candidate residual information and the structural correlation index with the standard dynamic range master image to determine whether to allow conditional gain information to enter the auxiliary data. The auxiliary data generation module is used to generate auxiliary data based on the decision result; when the decision is to allow, it generates auxiliary data containing physical baseline information, conditional gain information, and status flag fields; when the decision is to disallow, it generates auxiliary data containing physical baseline information and status flag fields. The encapsulation and synchronization module defines an adaptive auxiliary data structure to synchronously transmit auxiliary data and the standard dynamic range master to the projection end. The reconstruction module is used to reconstruct the high dynamic range signal based on auxiliary data. When conditional gain information exists, a multiplicative gain is applied to the high dynamic range baseline signal in the linear optical domain. When conditional gain information does not exist, the output is a high dynamic range signal that is equivalent to the high dynamic range baseline signal.

11. A method for distributing high dynamic range content, characterized in that, Applied to the production end of the video processing apparatus of claim 10, the method includes: Using the standard dynamic range master as the main asset, physical baseline information for global monotonic mapping is generated based on the standard dynamic range master; Generate candidate residual information of the reference high dynamic range image relative to the high dynamic range baseline signal corresponding to the physical baseline information; Semantic gating decisions are performed based on candidate residual information; wherein, the decision is based on the energy statistics index of the candidate residual information and the structural correlation index with the standard dynamic range master image to determine whether conditional gain information is allowed to enter the auxiliary data; Auxiliary data is generated based on the judgment result; when the judgment is allowed, auxiliary data containing physical baseline information, conditional gain information, and status flag fields is generated; when the judgment is disallowed, auxiliary data containing physical baseline information and status flag fields is generated.

12. An auxiliary data structure, characterized in that, The transmission end of the video processing apparatus of claim 10 includes: A parameterized lookup table encoded with physical baseline information representing a global monotonic mapping; When the status flag field indicates that conditional gain information is included, the auxiliary data structure includes a parameterized lookup table and conditional gain information; wherein, the conditional gain information is subjected to resolution scaling, includes region of interest block masking, quantization and compression encoding; When the status flag field indicates that conditional gain information is not included, the auxiliary data structure only includes a parameterized lookup table; The auxiliary data structure is encapsulated in an MXF container and establishes a time axis synchronization relationship with the standard dynamic range master through KLV encoding.

13. A method for displaying high dynamic range content, characterized in that, Applied to the projection end of the video processing apparatus of claim 10, the method includes: Receive standard dynamic range master and auxiliary data; When the status flag field indicates that conditional gain information is included, a multiplicative gain map is generated based on the conditional gain information. The nonlinear transformation of the standard dynamic range master is removed by inverse electro-optic conversion function to obtain a linear optical domain signal; A linear optical domain high dynamic range baseline signal is generated based on the linear optical domain signal and physical baseline information, and a pixel-by-pixel multiplicative operation is performed with the multiplicative gain map to obtain the linear optical domain high dynamic range signal; wherein, the multiplicative operation is a linear operator; The linear optical domain high dynamic range signal is converted into a high dynamic range signal by inverse electro-optic conversion function transformation. When the status flag field indicates that conditional gain information is not included, a high dynamic range baseline signal is generated based on the standard dynamic range master and physical baseline information. The high dynamic range signal is equivalent to the high dynamic range baseline signal.

Images