Pixtree Co., Ltd.

Pixtree, Inc.

Hyosong Kim

Sungul Ryoo

Hyeongduck Kim

Jaeseob Shin

Sehoon Son

Disclosed are a semantic image extrapolation method and a semantic image extrapolation apparatus. The present invention provides a technique for generating an empty region for image-extension in an image by using an extrapolated segmentation map and an inpainting technique. The present invention is to provide, considering that there is no information in an empty region for image-extension in an image, a semantic image extrapolation method, of first generating an extrapolated segmentation map on the basis of a segmentation map from an input image, and filling the empty region for image-extension in the image with information on the basis of the extrapolated segmentation map and the input image.

Semantic image extrapolation method and apparatus

The present embodiment provides an image restoration method and apparatus which generate independent different restoration models by performing learning for each of different resolutions, receive a distorted image, and apply a restoration model corresponding to the resolution of the distorted image among the independent different restoration models to restore the distorted image into an improved upscaled image centering on a restoration target object within the distorted image.

Image restoration method and apparatus

The present disclosure relates to a technical idea of compensating for degradation of an image quality during compression in an encoding process by using a parameter trained based on machine learning. An image processing apparatus includes an input frame buffer configured to store an input original frame, an encoding processor configured to encode the stored original frame, an output stream buffer configured to output the encoded original frame as an output stream for multiplexing, a multiplexing processor configured to control the output stream to be multiplexed and transmitted, a local decoding processor configured to output a restored stream by receiving the output stream from the output stream buffer and decoding the output stream, a restored frame buffer configured to store the output restored stream, and a machine learning processor configured to collect the input original frame and the output restored frame, output a parameter based on machine learning of the collected original frame and the collected restored frame, and control the multiplexing processor to transmit the output parameter to a decoding side.

Image processing apparatus and operation method for learning parameters based on machine learning

The present disclosure relates to a technical idea of compensating for degradation of an image quality during compression in an encoding process by using a parameter trained based on machine learning within a codec. An image processing apparatus includes an input frame buffer configured to store an input original frame, an encoding processor configured to encode the stored original frame, an output stream buffer configured to output the encoded original frame as an output stream for multiplexing, and a multiplexing processor configured to control the output stream to be multiplexed and transmitted. The encoding processor outputs a filter parameter trained based on machine learning using the stored original frame and a restored frame obtained by the encoding. A device on a decoding side decodes the output stream by using the output filter parameter.

Growing community of inventors

Seoul, South Korea

Hyosong Kim