A device includes a memory and one or more processors. The memory is configured to store instructions. The one or more processors are configured to execute the instructions to apply a neural network to a first image to generate an enhanced image. The one or more processors are also configured to execute the instructions to adjust at least a portion of a high-frequency component of the enhanced image based on a motion compensation operation to generate an adjusted high-frequency image component. The one or more processors are further configured to execute the instructions to combine a low-frequency component of the enhanced image and the adjusted high-frequency image component to generate an adjusted enhanced image.

An electronic device is provided. The electronic device includes a communication module, a camera module, a memory configured to store capability-related information about an external electronic device and information related to content, and a processor configured to be operatively connected to the communication module, the camera module, and the memory. The processor may generate at least one first content corresponding to a user using the camera module, may convert the at least one first content into at least one second content, based on the capability-related information about the external electronic device and the information related to the content, and may transmit the converted at least one second content to the external electronic device through the communication module.

An electronic device is provided. The electronic device includes a communication module, a camera module, a memory configured to store capability-related information about an external electronic device and information related to content, and a processor configured to be operatively connected to the communication module, the camera module, and the memory. The processor may generate at least one first content corresponding to a user using the camera module, may convert the at least one first content into at least one second content, based on the capability-related information about the external electronic device and the information related to the content, and may transmit the converted at least one second content to the external electronic device through the communication module.

A neural-network-based signal processing method and apparatus according to the present invention may: receive a bitstream including information about a neural network model, wherein the bitstream includes at least one neural network access unit; obtain information about the at least one neural network access unit from the bitstream; and reconstruct the neural network model on the basis of the information about the at least one neural network access unit.

A computer-implemented method of training an image generative network f.sub.θ for a set of training images, in which an output image {circumflex over (x)} is generated from an input image x of the set of training images non-losslessly, and in which a proxy network is trained for a gradient intractable perceptual metric that evaluates a quality of an output image {circumflex over (x)} given an input image x, the method of training using a plurality of scales for input images from the set of training images. In an embodiment, a blindspot network b.sub.α is trained which generates an output image {tilde over (x)} from an input image x. Related computer systems, computer program products and computer-implemented methods of training are disclosed.

There is disclosed an encoder, a decoder, related methods, and non-transitory storage units storing instructions which, when executed by a computer, cause the computer to perform the methods.

At an encoder (300), after a spatial transformation stage (304), there is obtained a spatially transformed version (306) of input image information (302) having multiple bands and, for each band, multiple transform band coefficients. After the generation of precincts (311), each comprising transform coefficients covering a predetermined spatial area of the input image information (302), there is provided a component transformation stage (320, 325), to apply one component transformation (CTr) selected (327) out of a plurality of predetermined component transformations, to each band (102′) of each precinct (311). Hence, there is obtained a spatially transformed and color transformed version (323) of the input image information (302), which is subsequently quantized and entropy encoded.

There is disclosed an encoder, a decoder, related methods, and non-transitory storage units storing instructions which, when executed by a computer, cause the computer to perform the methods.

At an encoder (300), after a spatial transformation stage (304), there is obtained a spatially transformed version (306) of input image information (302) having multiple bands and, for each band, multiple transform band coefficients. After the generation of precincts (311), each comprising transform coefficients covering a predetermined spatial area of the input image information (302), there is provided a component transformation stage (320, 325), to apply one component transformation (CTr) selected (327) out of a plurality of predetermined component transformations, to each band (102′) of each precinct (311). Hence, there is obtained a spatially transformed and color transformed version (323) of the input image information (302), which is subsequently quantized and entropy encoded.

A point cloud coding method, including: partitioning a point cloud, and determining a current coding block; determining a quantization parameter optimization enable identifier, a hierarchical level index, and a quantization parameter offset parameter of the current coding block; determining, according to the hierarchical level index and the quantization parameter offset parameter, a quantization step Qstep corresponding to a level of detail; and upon detecting that the quantization parameter optimization enable identifier is a first value, performing operations for one or more points included in the current coding block.

Provided are a device and operating method thereof for obtaining compression ratio information for recognizing a target object in an image using a deep neural network model, and compressing an image using the compression ratio information and encoding the compressed image. According to an embodiment of the present disclosure, there is provided a device that receives an image via at least one camera or a communication interface, obtains a feature map for detecting a target object in the received image, outputs a compression ratio for correctly recognizing the target object in the image by inputting the image and the feature map to a deep neural network model composed of pre-trained model parameters, and generates a bitstream by compressing the image using the output compression ratio and encoding the compressed image.

A three-dimensional data encoding method of encoding three-dimensional points includes: calculating a predicted value of attribute information of a first three-dimensional point in a prediction mode, using one or more items of attribute information of one or more second three-dimensional points in the vicinity of the first three-dimensional point; calculating a prediction residual that is a difference between the attribute information of the first three-dimensional point and the predicted value; and generating a bitstream including the prediction residual and prediction mode information indicating the prediction mode. The prediction mode is: one prediction mode among two or more prediction modes when a type of the attribute information of the first three-dimensional point is first attribute information including elements more than a predetermined threshold value; and one fixed prediction mode when the type is second attribute information including elements equal to or less than the predetermined threshold value.