Guangdong Oppo Mobile Telecommunications Corp., Ltd.

Zibo Meng

Chiuman Ho

Ming Chen

Runsheng Xu

Jin Pi

A method includes: aggregating information from a corresponding combination of all of multi-scale first dimensional receptive fields and each of multi-scale second dimensional receptive fields, so that information from multi-scale first and second dimensional receptive fields corresponding to the multi-scale second dimensional receptive fields is obtained; wherein the multi-scale first dimensional receptive fields being first dimensional is being one of spatial and temporal, and the multi-scale second dimensional receptive fields being second dimensional is being the other of spatial and temporal; wherein a corresponding first convolutional neural network operation set provides each of the multi-scale second dimensional receptive fields and each operation of the first convolutional neural network operation set has a corresponding first dimensional local to local constraint; and aggregating the information from the multi-scale first and second dimensional receptive fields to obtain aggregated multi-scale first and second dimensional receptive field information.

Method, system, and computer-readable medium for obtaining aggregated multi-scale first and second dimensional receptive field information

The present disclosure relates to a Temporal Information Aggregation (TIA) neural network block to extract underlying multiscale temporal information. By applying TIA, information in different temporal scales may be effectively extracted. The TIA block may be implemented as a block and thus may be inserted into any architectures. The extracted multi-scale temporal information contributes to the final output as a residual.

Video understanding neural network systems and methods using the same

An image processing method, an electronic device, and a computer-readable storage medium are provided. In the method, a first image is processed through a U-net to obtain a second image, and the second image is a noise map of the first image. The U-net includes an encoding network, a decoding network, and a bottleneck network between the encoding network and the decoding network. The bottleneck network includes a global pooling layer, a bilinear upscaling layer, and a 1×1 convolutional layer. Moreover, a third image is generated according to the first image and the second image, and the third image is a denoised map of the first image.

Image denoising method, device, and computer-readable medium using U-net

Provided are method and apparatus for image processing. A neural network includes an encoding network, an intermediate network, and a decoding network including multiple input layers and an output layer. In the method, at an input layer of the decoding network, first output data is received from a previous layer, and a first operation is performed on the first output data to obtain first input data of the input layer, the input layer is any one of multiple input layers. At the input layer, second output data is received from a corresponding layer of the encoding network, and a second operation is performed on the second output data to obtain second input data of the input layer. Output data of the input layer is obtained according to the first and second input data. Operations are performed in a next layer based on the output data to obtain an output image.

Method and apparatus for image processing

Provided are a method and apparatus for image processing and a terminal. In the method, an input image is received and processed in a neural network to obtain an output image according to global information of the input image. The terminal includes at least one processor; and a memory coupled with the at least one processor and configured to store instructions which, when executed by the at least one processor, are operable with the at least one processor to implement a neural network to receive an input image and process the input image in the neural network to obtain an output image according to global information of the input image.

Method and apparatus for image processing, terminal

In an embodiment, a method includes receiving and processing a first color image by an encoder. The first color image includes a first portion of the first color image and a second portion of the first color image located at different locations of the first color image. The encoder is configured to output at least one first feature map including fused global information and local information such that whether a color consistency relationship between the first portion of the first color image and the second portion of the first color image exists is encoded into the fused global information and local information.

Method, system, and computer-readable medium for improving color quality of images

A method includes receiving and processing a first image by an encoder-decoder network. The first image includes a first portion and a second portion located at different locations. The encoder-decoder network includes an encoder and a decoder. The encoder is configured to output at least one feature map including global information and local information such that whether a color consistency relationship between the first portion and the second portion of the first image exists is encoded into the global information and the local information. The decoder is configured to output a second image generated from the at least one feature map, wherein a first portion of the second image corresponding to the first portion of the first image and a second portion of the second image corresponding to the second portion of the first image are restored considering whether the color consistency relationship exists.

Method and terminal for improving color quality of images

In an embodiment, a method includes receiving a low-light digital image; generating, by at least one processor, a resulting digital image by processing the low-light digital image with an encoder-decoder neural network comprising a plurality of convolutional layers classified into a downsampling stage and an upsampling stage, and a multi-scale context aggregating block configured to aggregate multi-scale context information of the low-light digital image and employed between the downsampling stage and the upsampling stage; and outputting, by the at least one processor, the resulting digital image to an output device. In the upsampling stage, spatial resolution increases by using a bilinear interpolation operation performed before every few convolutional layers to speed up the inference time of the network.

Method, system, and computer-readable medium for improving quality of low-light images

In an embodiment, a method includes receiving a digital image; generating, by at least one processor, a resulting digital image by processing the digital image with an encoder-decoder neural network comprising a plurality of convolutional layers classified into a downsampling stage and an upsampling stage, and a multi-scale context aggregating block configured to aggregate multi-scale context information of the digital image and employed between the downsampling stage and the upsampling stage; and outputting, by the at least one processor, the resulting digital image to an output device. Generating the resulting digital image includes performing a pooling operation after every few convolutional layers at the downsampling stage to decrease spatial resolution; performing an upscaling operation before every few convolutional layers at the upsampling stage to increase the spatial resolution; and performing a channel-wise dropout operation following each of the convolutional layers at the downsampling stage and the upsampling stage.

In an embodiment, a method includes receiving a low-light digital image; generating, by at least one processor, a resulting digital image by processing the low-light digital image with an encoder-decoder neural network comprising a plurality of convolutional layers classified into a downsampling stage and an upscaling stage, and a multi-scale context aggregating block configured to aggregate multi-scale context information of the low-light digital image and employed between the downsampling stage and the upscaling stage; and outputting, by the at least one processor, the resulting digital image to an output device.

Growing community of inventors

Palo Alto, CA, United States of America

Zibo Meng