A system includes a processing unit communicatively coupled to a detector array of an optical wafer characterization system. The processing unit is configured to perform one or more steps of a method or process including the steps of acquiring one or more target images of a target location on a wafer from the detector array, applying a de-noising filter to at least the one or more target images, determining one or more difference images from one or more reference images and the one or more target images, and up-sampling the one or more difference images to generate one or more up-sampled images. One or more wafer defects are detectable in the one or more difference images or the up-sampled images.

According to one embodiment, a medical information processing method generates a high-contrast image by applying conversion processing to a first medical image captured by a first diagnostic apparatus and having a first contrast in a region of interest, the high-contrast image having a contrast higher than a contrast of a second medical image obtained by a second diagnostic apparatus. The method generates a pseudo second medical image by applying image processing to the high-contrast image, the pseudo second medical image simulating the second medical image and having a second contrast lower than the first contrast. The method trains a model using the pseudo second medical image as input data and the high-contrast image as ground truth data, and generates a trained model.

Examples of systems and methods for augmented facial animation are generally described herein. A method for mapping facial expressions to an alternative avatar expression may include capturing a series of images of a face, and detecting a sequence of facial expressions of the face from the series of images. The method may include determining an alternative avatar expression mapped to the sequence of facial expressions, and animating an avatar using the alternative avatar expression.

An image processing device (100) includes a noise reducer (22) including a pixel ratio acquirer (23) configured to acquire a pixel value ratio (α) between a scattered-ray reduced image (52) after reduction of a scattered-ray component and a radiation image (51) before the reduction of the scattered-ray component, the noise reducer being configured to reduce a noise component from the scattered-ray reduced image based on the pixel value ratio.

A neural network architecture is disclosed for restoring noisy data. The neural network is a blind-spot network that can be trained according to a self-supervised framework. In an embodiment, the blind-spot network includes a plurality of network branches. Each network branch processes a version of the input data using one or more layers associated with kernels that have a receptive field that extends in a particular half-plane relative to the output value. In one embodiment, the versions of the input data are offset in a particular direction and the convolution kernels are rotated to correspond to the particular direction of the associated network branch. In another embodiment, the versions of the input data are rotated and the convolution kernel is the same for each network branch. The outputs of the network branches are composited to de-noise the image. In some embodiments, Bayesian filtering is performed to de-noise the input data.

An apparatus, system and method for regulating fluid flow are disclosed. The apparatus includes a flow rate sensor and a valve. The flow rate sensor uses images to estimate flow through a drip chamber and then controls the valve based on the estimated flow rate. The valve comprises a rigid housing disposed around the tube in which fluid flow is being controlled. Increasing the pressure in the housing controls the size of the lumen within the tube by deforming the tube, therefore controlling flow through the tube.

System and methods are disclosed for capturing license plate (LP) information of a vehicle in relative motion to a camera device. In one example, the camera system detects the LP in multiple frames, then aligns and geometrically rectifies the image of the LP by scaling, warping, rotating, and/or performing other functions on the images. The camera system may optimize capturing of the LP information by executing a temporal noise filter on the aligned, geometrically rectified images to generate a composite image of the LP for optical character recognition. In some examples, the camera device may include an image sensor, such as a high dynamic range (HDR) sensor, modified to set long and short exposures of the HDR sensor to capture frames of a vehicle's LP, but without consolidating the images into a composite image. The camera system may set optimal exposure settings based on detected relative speed of the vehicle.

A video coding device may be configured to perform directional Bi-directional optical flow (BDOF) refinement on a coding unit (CU). The device may determine the direction in which to perform directional BDOF refinement. The device may calculate the vertical direction gradient difference and the horizontal direction gradient difference for the CU. The vertical direction gradient difference may indicate the difference between the vertical gradients for a first reference picture and the vertical gradients for a second reference picture. The horizontal direction gradient difference may indicate the difference between the horizontal gradients for the first reference picture and the horizontal gradients for the second reference picture. The video coding device may determine the direction in which to perform directional BDOF refinement based on the vertical direction gradient difference and the horizontal direction gradient difference. The video coding device may perform directional BDOF refinement in the determined direction.

Apparatuses, systems, and techniques are presented to train and utilize one or more neural networks. A denoising diffusion generative adversarial network (denoising diffusion GAN) reduces a number of denoising steps during a reverse process. The denoising diffusion GAN does not assume a Gaussian distribution for large steps of the denoising process and applies a multi-model model to permit denoising with fewer steps. Systems and methods further minimize a divergence between a diffused real data distribution and a diffused generator distribution over several timesteps. Accordingly, various embodiments may enable faster sample generation, in which the samples are generated from noise using the denoising diffusion GAN.

One embodiment provides a method comprising receiving an input video comprising low-resolution (LR) frames and corresponding super-resolution (SR) frames, and generating a motion-compensated previous SR frame based on a current LR frame of the video and a motion-compensated previous residual frame of the video. The previous SR frame aligns with a current SR frame corresponding to the current LR frame. The method further comprises, in response to determining there is a mismatch between the previous SR frame and the current SR frame, correcting in the current SR frame errors that result from motion compensation based on the motion-compensated previous SR frame. The method further comprises restoring details to the current SR frame that were lost as a result of the correcting, and suppressing flickers of the current SR frame on the frequency domain, resulting in a flicker-suppressed current SR frame for presentation on a display.