The embodiments of the application provide an image super-resolution reconstruction method and device featuring labeled data sharpening. The method comprises: establishing a deep neural network for image reconstruction from low resolution to high resolution, obtaining low-resolution sample data by sub-sampling and Sobel operator calculation, and then sharpening labeled data to a certain extent so that edges of the labeled data are clearer and object features are easier to see. A model is obtained by training the deep neural network, and the model is used to perform super-resolution processing on the image to obtain an image with a higher resolution. The obtained image has a clear texture, the resolution and definition are significantly improved, and the quality evaluation indexes of the image are better.

The present disclosure relates to systems, methods, and non-transitory computer-readable media that upsample and refine segmentation masks. Indeed, in one or more implementations, a segmentation mask refinement and upsampling system upsamples a preliminary segmentation mask utilizing a patch-based refinement process to generate a patch-based refined segmentation mask. The segmentation mask refinement and upsampling system then fuses the patch-based refined segmentation mask with an upsampled version of the preliminary segmentation mask. By fusing the patch-based refined segmentation mask with the upsampled preliminary segmentation mask, the segmentation mask refinement and upsampling system maintains a global perspective and helps avoid artifacts due to the local patch-based refinement process.

The present disclosure discloses an image super-resolution processing method and device. A plurality of image processing engines with different resolutions are provided, and the image processing engines include a general engine and a special engine. The method includes: obtaining a to-be-processed target image, and determining whether a target engine matched with the target image exists in the general engine according to a resolution of the target image; in response to the target engine exists, selecting the target engine to perform super-resolution processing on the target image; and in response to the target engine does not exist, preprocessing the resolution of the target image into an input resolution supported by the special engine, and selecting the special engine to perform super-resolution processing on the preprocessed target image.

An image processing device is provided, which includes an image capture circuit and a processor. The image capturing circuit is configured for capturing a high-resolution image. The processor is connected to the image capturing circuit, and performing a super-resolution model and an attention model, where the processor is configured to perform following operations for: performing down sampling processing on the high-resolution image to generate a low-resolution image; performing super-resolution processing on the low-resolution image using the super-resolution model to generate a super-resolution image; applying the attention model to the high-resolution image and the super-resolution image to generate an attention weighted high-resolution image and an attention weighted super-resolution image, and calculating a first loss according to the attention weighted high-resolution image and the attention weighted super-resolution image, thereby updating the super-resolution model.

An image processing apparatus configured to achieve high definition of an image of a second image group by using a first image group, the second image group having less high-frequency components in a corresponding frame than the first image group, the image processing apparatus comprises a selection unit configured to select, based on a high-definition target image selected from the second image group, a pair of supervisory data to be used for learning, a learning model generation unit configured to generate a learning model by using the pair of supervisory data, an inference unit configured to infer high-frequency components of the high-definition target image by using the learning model generated, and an image generation unit configured to generate a high-definition image based on the high-definition target image and the high-frequency components.

An image sensor according to an embodiment of the present invention includes: a pixel array in which a plurality of pixels are arrayed in a grid shape, and which converts reflection light signals reflected from an object into electrical signals; an image processor which converts the electrical signals to generate subframes, and extracts pieces of second depth information having a higher resolution than pieces of first depth information extracted from a plurality of the subframes; and a memory for storing the pieces of first depth information, wherein the reflection light signals are input to the pixel array through mutually different optical paths shifted in sub-pixel units of the pixel array, and the memory stores a plurality of the pieces of first depth information that correspond to the mutually different optical paths.

A system for generating depth information from low-resolution images is configured to access a plurality of image frames capturing an environment, identify a first group of image frames from the plurality of image frames, and generate a first image comprising a first composite image of the environment using the first group of image frames as input. The first composite image has an image resolution that is higher than an image resolution of the image frames of the first group of image frames. The system is also configured to obtain a second image of the environment, where parallax exists between a capture perspective associated with the first image and a capture perspective associated with the second image. The system is also configured to generate depth information for the environment based on the first image and the second image.

The present application provides methods, devices, systems and non-transitory computer readable storage media for image processing. In an aspect, there is provided a computer-implemented method of image processing, the method comprising: receiving a low resolution image of an object; generating an edge map of the low resolution image by an edge extractor; and inputting the edge map and the low resolution image to a neural network to reconstruct a super resolution image of the object, wherein the neural network is trained using a multicomponent loss function.

Provided are an upscaling device, an upscaling method, and an upscaling program capable of keying processing at low cost and with high accuracy. An upscaling device includes an image obtaining section configured to obtain a high-resolution input image illustrating a background and a foreground, a low-resolution input image obtained by converting the high-resolution input image into a low-resolution image, and a low-resolution alpha image illustrating a region of the foreground in the low-resolution input image; a correlation obtaining section configured to obtain correlation between grayscale information of the low-resolution input image and an alpha value of the low-resolution alpha image for each corresponding pixel pair of the low-resolution input image and the low-resolution alpha image, and upscale the correlation into a resolution of the high-resolution input image; and an output section configured to output a high-resolution alpha image having a resolution same as that of the high-resolution input image on a basis of the high-resolution input image and the upscaled correlation.

A remote support system remotely supports an operation of a moving body based on an image captured by a camera installed on the moving body. The remote support system communicates with the moving body to receive the image captured by the camera. The remote support system determines, based on an input image that is the received image or a corrected image acquired by correcting the received image, whether or not congestion control that reduces an image quality of the image is performed in the moving body. When the congestion control is performed in the moving body, the remote support system apples a super-resolution technique to the input image to improve an image quality of the input image to generate an improved image. Then, the remote support system displays the improved image on a display device.