A method and/or device for predicting a compression quality of an image during image correction (e.g., image quality enhancement) in an electronic device, and/or processing the image, based on at least the prediction, may be provided. The electronic device may include a display module, a memory, and a processor, wherein the processor may operate to display an image through the display module, extract designated multiple blocks from the image in a designated scheme, estimate confidence for each of the multiple blocks, identify, based on the estimation of the confidence, a first block corresponding to an outlier to be excluded in quality prediction, and a second block for which quality prediction is possible, among the multiple blocks, exclude the first block among the multiple blocks from a subject of quality prediction, and classify a compression quality of the image by using at least the second block remaining after excluding the first block from among the multiple blocks.

A method for generating a subtraction image for digital subtraction angiography to reduce noise and movement artifacts. Obtaining a plurality of mask images of an object takes place before administering a contrast agent into the object and obtaining a map of the object after administering a contrast agent into the object. A first sum image is obtained from the plurality of mask images in that the plurality of mask images is summed in each case multiplied by an individual weighting. The individual weightings for each of the plurality of mask images are automatically determined by an optimization method, and the subtraction image is ascertained by subtraction of the sum image from the map.

In various examples, high beam control for vehicles may be automated using a deep neural network (DNN) that processes sensor data received from vehicle sensors. The DNN may process the sensor data to output pixel-level semantic segmentation masks in order to differentiate actionable objects (e.g., vehicles with front or back lights lit, bicyclists, or pedestrians) from other objects (e.g., parked vehicles). Resulting segmentation masks output by the DNN(s), when combined with one or more post processing steps, may be used to generate masks for automated high beam on/off activation and/or dimming or shading—thereby providing additional illumination of an environment for the driver while controlling downstream effects of high beam glare for active vehicles.

A method of removing raindrops from video images is provided. The method includes the steps of: training a raindrop image recognition model using a plurality raindrop training images labeled in a plurality of rainy-scene images; recognizing a plurality of raindrop images from a plurality of scene images in a video sequence using the raindrop image recognition model; and in response to a specific raindrop image in a current scene image satisfying a predetermined condition, replacing the specific raindrop image in the current scene image with an image region corresponding to the specific raindrop image in a specific scene image prior to the current scene image to generate an output scene image.

Methods and systems are provided for improving image quality of ultrasound images by automatically determining one or more image quality parameters via a plurality of separate image quality models. In one example, a method for an ultrasound system includes determining a plurality of image quality parameters of an ultrasound image acquired with the ultrasound system, each image quality parameter determined based on output from a separate image quality model, and outputting feedback to a user of the ultrasound system based on the plurality of image quality parameters.

There is provided an image processing device including a light sensor and a processor. The light sensor is used to detect light and output an image frame. The processor identifies intensity of ambient light according to an image parameter associated with the image frame. When the ambient light is identified to be strong enough, the processor performs an object identification directly using the image frame. When the ambient light is identified to be not enough, the processor firstly converts the image frame to a converted image using a machine learning model, and then performs the object identification using the converted image.

A modular architecture is provided for denoising Monte Carlo renderings using neural networks. The temporal approach extracts and combines feature representations from neighboring frames rather than building a temporal context using recurrent connections. A multiscale architecture includes separate single-frame or temporal denoising modules for individual scales, and one or more scale compositor neural networks configured to adaptively blend individual scales. An error-predicting module is configured to produce adaptive sampling maps for a renderer to achieve more uniform residual noise distribution. An asymmetric loss function may be used for training the neural networks, which can provide control over the variance-bias trade-off during denoising.

A method of depth estimation, including, receiving an image frame, determining a relative depth map based on the image frame, receiving a sparse depth frame, preprocessing the sparse depth frame, determining a scale-adjusted relative depth map based on the relative depth map and the preprocessed sparse depth frame and fusing the relative depth map and the scale-adjusted relative depth map to produce an absolute depth map.

In one embodiment, a computing system may access a video including a first frame and a second frame. The computing system may determine first sampling locations for the first frame and determine second sampling locations for the second frame by transforming the first sampling locations to the second frame according to an optical flow between the first frame and the second frame. The computing system may detect one or more invalid second sampling locations based on determining pixels in the first frame corresponding to the first sampling locations do not match pixels in the second frame corresponding to the second sampling locations. The computing system may reject the one or more invalid second sampling locations to determine third sampling locations for the second frame. The computing system may generate a sample of the video.

Examples are provided relating to recovering depth data from noisy phase data of low-signal pixels. One example provides a computing system, comprising a logic machine, and a storage machine holding instructions executable by the logic machine to process depth data by obtaining depth image data and active brightness image data for a plurality of pixels, the depth image data comprising phase data for a plurality of frequencies, and identifying low-signal pixels based at least on the active brightness image data. The instructions are further executable to apply a denoising filter to phase data of the low-signal pixels to obtain denoised phase data and not applying the denoising filter to phase data of other pixels. The instructions are further executable to, after applying the denoising filter, perform phase unwrapping on the phase data for the plurality of frequencies to obtain a depth image, and output the depth image.