An image coding device is provided with a determination unit which determines whether to apply an orthogonal transform to a transform block obtained by dividing a prediction difference signal indicating a difference between an input image and a predicted image or perform a transform skip by which the orthogonal transform is not applied, and an orthogonal transform unit which performs processing selected on the basis of the determination, the image coding device comprising a quantization unit which, when the transform skip is selected on the basis of the determination, quantizes the transform block using a first quantization matrix in which the quantization roughnesses of all elements previously shared with a decoding side are equal, and when the orthogonal transform is applied to the transform block on the basis of the determination, quantizes the transform block using the first quantization matrix or a second quantization matrix that is transmitted to the decoding side.

An image coding device is provided with a determination unit which determines whether to apply an orthogonal transform to a transform block obtained by dividing a prediction difference signal indicating a difference between an input image and a predicted image or perform a transform skip by which the orthogonal transform is not applied, and an orthogonal transform unit which performs processing selected on the basis of the determination, the image coding device comprising a quantization unit which, when the transform skip is selected on the basis of the determination, quantizes the transform block using a first quantization matrix in which the quantization roughnesses of all elements previously shared with a decoding side are equal, and when the orthogonal transform is applied to the transform block on the basis of the determination, quantizes the transform block using the first quantization matrix or a second quantization matrix that is transmitted to the decoding side.

In one embodiment, an apparatus comprises a memory and a processor. The memory is to store visual data associated with a visual representation captured by one or more sensors. The processor is to: obtain the visual data associated with the visual representation captured by the one or more sensors, wherein the visual data comprises uncompressed visual data or compressed visual data; process the visual data using a convolutional neural network (CNN), wherein the CNN comprises a plurality of layers, wherein the plurality of layers comprises a plurality of filters, and wherein the plurality of filters comprises one or more pixel-domain filters to perform processing associated with uncompressed data and one or more compressed-domain filters to perform processing associated with compressed data; and classify the visual data based on an output of the CNN.

A process for selecting a transform set for a prediction block. The process can be used in both an encoder and a decoder. For example, the process can be used in both an encoder and a decoder for a prediction block that has been predicted from a reference block. In some embodiments, both the prediction block and the reference block are intra blocks.

There is disclosed a method of encoding an input signal, the method comprising: receiving a base encoded signal, the base encoded signal being generated by feeding an encoder with a down-sampled version of an input signal; producing a first residual signal by: decoding the base encoded signal to produce a first decoded signal; and using a difference between the base decoded signal and the down-sampled version of the input signal to produce the first residual signal; producing a second residual signal by: correcting the base decoded signal using the residual signal to create a corrected decoded version; up-sampling the corrected decoded version; and using a difference between the up-sampled corrected decoded signal and the input signal to produce the second residual signal; wherein the up-sampling is one of bilinear or bicubic up-sampling. A corresponding decoding method is also disclosed.

There is disclosed a method of encoding an input signal, the method comprising: receiving a base encoded signal, the base encoded signal being generated by feeding an encoder with a down-sampled version of an input signal; producing a first residual signal by: decoding the base encoded signal to produce a first decoded signal; and using a difference between the base decoded signal and the down-sampled version of the input signal to produce the first residual signal; producing a second residual signal by: correcting the base decoded signal using the residual signal to create a corrected decoded version; up-sampling the corrected decoded version; and using a difference between the up-sampled corrected decoded signal and the input signal to produce the second residual signal; wherein the up-sampling is one of bilinear or bicubic up-sampling. A corresponding decoding method is also disclosed.

A computing device performs a method of decoding video data by reconstructing a luma block corresponding to a chroma block; searching a sub-group of a plurality of reconstructed neighboring luma samples in a predefined order to identify a maximum luma sample and a minimum luma sample; computing a down-sampled maximum luma sample corresponding to the maximum luma sample; computing a down-sampled minimum luma sample corresponding to the minimum luma sample; generating a linear model using the down-sampled maximum luma sample, the down-sampled minimum luma sample, the first reconstructed chroma sample, and the second reconstructed chroma sample; computing down-sampled luma samples from luma samples of the reconstructed luma block, wherein each down-sampled luma sample corresponds to a chroma sample of the chroma block; and predicting chroma samples of the chroma block by applying the liner model to the corresponding down-sampled luma samples.

A computing device performs a method of decoding video data by reconstructing a luma block corresponding to a chroma block; searching a sub-group of a plurality of reconstructed neighboring luma samples in a predefined order to identify a maximum luma sample and a minimum luma sample; computing a down-sampled maximum luma sample corresponding to the maximum luma sample; computing a down-sampled minimum luma sample corresponding to the minimum luma sample; generating a linear model using the down-sampled maximum luma sample, the down-sampled minimum luma sample, the first reconstructed chroma sample, and the second reconstructed chroma sample; computing down-sampled luma samples from luma samples of the reconstructed luma block, wherein each down-sampled luma sample corresponds to a chroma sample of the chroma block; and predicting chroma samples of the chroma block by applying the liner model to the corresponding down-sampled luma samples.

A method is provided for processing a video bitstream which includes a first picture and a sequence parameter set containing information indicating a mode is enabled in which a first area of a picture can be extracted and replaced by a second area of another picture. Using the mode, the first picture is modified by extracting the first area and replacing the extracted area with the second area from the other picture. The modified first picture is stored and a predictor is formed using the modified first picture as a reference picture. A block of a subsequent picture is decoded using the predictor.

A method is provided for processing a video bitstream which includes a first picture and a sequence parameter set containing information indicating a mode is enabled in which a first area of a picture can be extracted and replaced by a second area of another picture. Using the mode, the first picture is modified by extracting the first area and replacing the extracted area with the second area from the other picture. The modified first picture is stored and a predictor is formed using the modified first picture as a reference picture. A block of a subsequent picture is decoded using the predictor.