Abstract: Different candidate image data feature types are evaluated to identify one or more specific image data feature types to be used in training a prediction model for optimizing one or more image metadata parameters. A plurality of image data features of the one or more selected image data feature types is extracted from one or more images. The plurality of image data features of the one or more selected image data feature types is reduced into a plurality of significant image data features. A total number of image data features in the plurality of significant image data features is no larger than a total number of image data features in the plurality of image data features of the one or more selected image data feature types. The plurality of significant image data features is applied to training the prediction model for optimizing one or more image metadata parameters.

Abstract: Methods and systems for generating an interpolated reshaping function for the efficient coding of high-dynamic range images are provided. The interpolated reshaping function is constructed based on a set of pre-computed basis reshaping functions. Interpolation schemes are derived for pre-computed basis reshaping functions represented as look-up tables, multi-segment polynomials, or matrices of coefficients in a multivariate, multi-regression representation. Encoders and decoders using asymmetric reshaping and interpolated reshaping functions for mobile applications are also presented.

Abstract: 3D mapping statistics are generated for a first image of a first dynamic range and a second image of a second dynamic range different from the first dynamic range. Multivariate multiple regression (MMR) coefficients are generated by solving an optimization problem formulated using an MMR matrix built with the 3D mapping statistics without a letterbox constraint, and used to generate chroma mappings for predicting chroma codeword values of the second image. It is determined whether a letterbox exists in the images. If so, it is determined whether the chroma mappings accurately predict chroma codeword values in the second image. A reconstructed image generated by a recipient device by backward reshaping one of the images is rendered by a display device operating in conjunction with the recipient device.

Abstract: Training image pairs comprising training SDR image and corresponding training HDR images are received. Each training image pair in the training image pairs comprises a training SDR image and a corresponding training HDR image. The training SDR image and the corresponding training HDR image in the training image pair depict same visual content but with different luminance dynamic ranges. Training image feature vectors are extracted from training SDR images in the training image pairs. The training image feature vectors are used to train backward reshaping metadata prediction models for predicting operational parameter values of backward reshaping mappings used to backward reshape SDR images into mapped HDR images.

Abstract: Methods and systems for generating an interpolated reshaping function for the efficient coding of high-dynamic range images are provided. The interpolated reshaping function is constructed based on a set of pre-computed basis reshaping functions. Interpolation schemes are derived for pre-computed basis reshaping functions represented as look-up tables, multi-segment polynomials, or matrices of coefficients in a multivariate, multi-regression representation. Encoders and decoders using asymmetric reshaping and interpolated reshaping functions for mobile applications are also presented.

Abstract: Different candidate image data feature types are evaluated to identify one or more specific image data feature types to be used in training a prediction model for optimizing one or more image metadata parameters. A plurality of image data features of the one or more selected image data feature types is extracted from one or more images. The plurality of image data features of the one or more selected image data feature types is reduced into a plurality of significant image data features. A total number of image data features in the plurality of significant image data features is no larger than a total number of image data features in the plurality of image data features of the one or more selected image data feature types. The plurality of significant image data features is applied to training the prediction model for optimizing one or more image metadata parameters.

Abstract: 3D mapping statistics are generated for a first image of a first dynamic range and a second image of a second dynamic range different from the first dynamic range. Multivariate multiple regression (MMR) coefficients are generated by solving an optimization problem formulated using an MMR matrix built with the 3D mapping statistics without a letterbox constraint, and used to generate chroma mappings for predicting chroma codeword values of the second image. It is determined whether a letterbox exists in the images. If so, it is determined whether the chroma mappings accurately predict chroma codeword values in the second image. A reconstructed image generated by a recipient device by backward reshaping one of the images is rendered by a display device operating in conjunction with the recipient device.

Abstract: A tone-mapping function that maps input images of a high dynamic range into reference tone-mapped images of a relatively narrow dynamic range is generated. A luma forward reshaping function is derived, based on first bit depths and second bit depths, for forward reshaping luma codewords of the input images into forward reshaped luma codewords of forward reshaped images approximating the reference tone-mapped images. A chroma forward reshaping mapping is derived for predicting chroma codewords of the forward reshaped images. Backward reshaping metadata that is to be used by recipient devices to generate a luma backward reshaping function and a chroma backward reshaping mapping is transmitted with the forward reshaped images to the recipient devices. Techniques for the joint derivation of forward luma and chroma reshaping functions are also presented.

Abstract: Real-time forward reshaping, comprising selecting a statistical sliding window that indexes with the current frame, having also, a look-back frame and a look-ahead frame, determining whether they are part of the current scene, determining a noise parameter, a luma transfer function and a luma forward reshaping function based on the luma transfer function and the noise parameter within the current scene, selecting a central tendency sliding window of the current frame and the look-back frame within the current scene, and determining a central tendency luma forward reshaping function. The chroma reshaping comprises analyzing statistics for the extended dynamic range (EDR) weights and EDR upper bounds, mapping these to standard dynamic range (SDR) weights and SDR upper bounds based on the central tendency luma forward reshaping function, determining a chroma content-dependent polynomial and a central tendency chroma forward reshaping polynomial and generating chroma MMR coefficients.

Abstract: In a method to reconstruct a high dynamic range video signal, a decoder receives parameters in the input bitstream to generate a prediction function. Using the prediction function, it generates a first set of nodes for a first prediction lookup table, wherein each node is characterized by an input node value and an output node value. Then, it modifies the output node values of one or more of the first set of nodes to generate a second set of nodes for a second prediction lookup table, and generates output prediction values using the second lookup table. Low-complexity methods to modify the output node value of a current node in the first set of nodes based on computing modified slopes between the current node and nodes surrounds the current node are presented.

Abstract: A standard dynamic range (SDR) image is received. Composer metadata of the first level through the N-th level is generated. Composer metadata of the j-th level is generated based on the composer metadata of the first level through (j?1)-th level. The composer metadata of the first level through the composer metadata of the j-th level is to be used for mapping the SDR image to the j-th target image specifically optimized for the j-th reference target display. The SDR image is encoded with the composer metadata of the first level through the k-th level in an output SDR video signal, where 1<=k<=N. A display device renders a display image derived from a composed target image composed from the SDR image based on the composer metadata of the first level through the k-th level in the output SDR video signal.

Abstract: Given HDR and SDR video inputs representing the same content, segment-based methods are described to generate a backward-compatible reshaped SDR video which preserves the artistic intent or “look” of the inputs and satisfies other coding requirements. For each frame in a segment, reshaping functions are generated based on a support frames set determined based on a sliding window of frames that is adjusted based on scene cuts in the segment and which may include frames from both the current segment and neighboring segments. For luma reshaping, a mapping that preserves the cumulative density function of the luminance histogram values in the EDR and SDR inputs is combined with a minimum codeword allocation derived based on the EDR signal and the support frame set. For chroma reshaping, methods for segment-based forward and backward reshaping using multivariate, multi-regression models are also presented.

Abstract: In a method to reconstruct a high dynamic range video signal, a decoder receives parameters in the input bitstream to generate a prediction function. Using the prediction function, it generates a first set of nodes for a first prediction lookup table, wherein each node is characterized by an input node value and an output node value. Then, it modifies the output node values of one or more of the first set of nodes to generate a second set of nodes for a second prediction lookup table, and generates output prediction values using the second lookup table. Low-complexity methods to modify the output node value of a current node in the first set of nodes based on computing modified slopes between the current node and nodes surrounding the current node are presented.

Abstract: Given HDR and SDR video inputs representing the same content, segment-based methods are described to generate a backward-compatible reshaped SDR video which preserves the artistic intent or “look” of the inputs and satisfies other coding requirements. For each frame in a segment, reshaping functions are generated based on a support frames set determined based on a sliding window of frames that is adjusted based on scene cuts in the segment and which may include frames from both the current segment and neighboring segments. For luma reshaping, a mapping that preserves the cumulative density function of the luminance histogram values in the EDR and SDR inputs is combined with a minimum codeword allocation derived based on the EDR signal and the support frame set. For chroma reshaping, methods for segment-based forward and backward reshaping using multivariate, multi-regression models are also presented.

Abstract: A standard dynamic range (SDR) image is received. Composer metadata of the first level through the N-th level is generated. Composer metadata of the j-th level is generated based on the composer metadata of the first level through (j?1)-th level. The composer metadata of the first level through the composer metadata of the j-th level is to be used for mapping the SDR image to the j-th target image specifically optimized for the j-th reference target display. The SDR image is encoded with the composer metadata of the first level through the k-th level in an output SDR video signal, where 1<=k<=N. A display device renders a display image derived from a composed target image composed from the SDR image based on the composer metadata of the first level through the k-th level in the output SDR video signal.

Abstract: In a system for coding high dynamic range (HDR) images using lower-dynamic range (LDR) images, a reshaping function allows for a more efficient distribution of the codewords in the lower dynamic range images for improved compression. A trim pass of the LDR images by a colorist may satisfy a director's intent for a given “look,” but may also result in unpleasant clipping artifacts in the reconstructed HDR images. Given an original forward reshaping function which maps HDR luminance values to LDR pixel values, a processor identifies areas of potential clipping and generates modified forward and backward reshaping functions to reduce the visibility of potential artifacts from the trim pass process while preserving the director's intent.

Abstract: Real-time forward reshaping, comprising selecting a statistical sliding window that indexes with the current frame, having also, a look-back frame and a look-ahead frame, determining whether they are part of the current scene, determining a noise parameter, a luma transfer function and a luma forward reshaping function based on the luma transfer function and the noise parameter within the current scene, selecting a central tendency sliding window of the current frame and the look-back frame within the current scene, and determining a central tendency luma forward reshaping function. The chroma reshaping comprises analyzing statistics for the extended dynamic range (EDR) weights and EDR upper bounds, mapping these to standard dynamic range (SDR) weights and SDR upper bounds based on the central tendency luma forward reshaping function, determining a chroma content-dependent polynomial and a central tendency chroma forward reshaping polynomial and generating chroma MMR coefficients.

Abstract: In a method to reconstruct a high dynamic range video signal, a decoder receives parameters in the input bitstream to generate a prediction function. Using the prediction function, it generates a first set of nodes for a first prediction lookup table, wherein each node is characterized by an input node value and an output node value. Then, it modifies the output node values of one or more of the first set of nodes to generate a second set of nodes for a second prediction lookup table, and generates output prediction values using the second lookup table. Low-complexity methods to modify the output node value of a current node in the first set of nodes based on computing modified slopes between the current node and nodes surrounds the current node are presented.

Abstract: An SDR CDF is constructed based on an SDR histogram generated from a distribution of SDR codewords in SDR images. An HDR CDF is constructed based on an HDR histogram generated from a distribution of HDR codewords in HDR images that correspond to the SDR images. A histogram transfer function is generated based on the SDR CDF and the HDR CDF. The SDR images are transmitted along with backward reshaping metadata to recipient devices. The backward reshaping metadata is generated at least in part on the histogram transfer function.

Abstract: A tone-mapping function that maps input images of a high dynamic range into reference tone-mapped images of a relatively narrow dynamic range is generated. A luma forward reshaping function is derived, based on first bit depths and second bit depths, for forward reshaping luma codewords of the input images into forward reshaped luma codewords of forward reshaped images approximating the reference tone-mapped images. A chroma forward reshaping mapping is derived for predicting chroma codewords of the forward reshaped images. Backward reshaping metadata that is to be used by recipient devices to generate a luma backward reshaping function and a chroma backward reshaping mapping is transmitted with the forward reshaped images to the recipient devices. Techniques for the joint derivation of forward luma and chroma reshaping functions are also presented.