Samsung Electronics Co., Ltd.

Texas Instruments Corporation

Yingmao Li

Chenchi Luo

Youngjun Yoo

Hamid Rahim Sheikh

John Seokjun Lee

Gyeongmin Choe

Michael Oliver Polley

Kaimo Lin

George Q Chen

Kushal Kardam Vyas

Youngmin Kim

David D Liu

Vikram Vijayanbabu Appia

Zeeshan Nadir

Tarek Aziz Lahlou

Ziguo Zhong

Jun Ki Cho

Seung-Chul Jeon

An apparatus includes at least one processing device configured to obtain input frames from a video. The at least one processing device is also configured to generate a forward flow from a first input frame to a second input frame and a backward flow from the second input frame to the first input frame. The at least one processing device is further configured to generate an occlusion map at an interpolated frame coordinate using the forward flow and the backward flow. The at least one processing device is also configured to generate a consistency map at the interpolated frame coordinate using the forward flow and the backward flow. In addition, the at least one processing device is configured to perform blending using the occlusion map and the consistency map to generate an interpolated frame at the interpolated frame coordinate.

Pixel blending for synthesizing video frames with occlusion and watermark handling

A method includes obtaining multiple video frames. The method also includes determining whether a bi-directional optical flow between the multiple video frames satisfies an image quality criterion for bi-directional consistency. The method further includes identifying a non-linear curve based on pixel coordinate values from at least two of the video frames. The at least two video frames include first and second video frames. The method also includes generating interpolated video frames between the first and second video frames by applying non-linear interpolation based on the non-linear curve. In addition, the method includes outputting the interpolated video frames for presentation.

Accurate optical flow interpolation optimizing bi-directional consistency and temporal smoothness

A method includes obtaining a first optical flow vector representing motion between consecutive video frames during a previous time step. The method also includes generating a first predicted optical flow vector from the first optical flow vector using a trained prediction model, where the first predicted optical flow vector represents predicted motion during a current time step. The method further includes refining the first predicted optical flow vector using a trained update model to generate a second optical flow vector representing motion during the current time step. The trained update model uses the first predicted optical flow vector, a video frame of the previous time step, and a video frame of the current time step to generate the second optical flow vector.

Multi-frame optical flow network with lossless pyramid micro-architecture

A method includes obtaining, using a stationary sensor of an electronic device, multiple image frames including first and second image frames. The method also includes generating, using multiple previously generated motion vectors, a first motion-distorted image frame using the first image frame and a second motion-distorted image frame using the second image frame. The method further includes adding noise to the motion-distorted image frames to generate first and second noisy motion-distorted image frames. The method also includes performing (i) a first multi-frame processing (MFP) operation to generate a ground truth image using the motion-distorted image frames and (ii) a second MFP operation to generate an input image using the noisy motion-distorted image frames. In addition, the method includes storing the ground truth and input images as an image pair for training an artificial intelligence/machine learning (AI/ML)-based image processing operation for removing image distortions caused by handheld image capture.

Hand motion pattern modeling and motion blur synthesizing techniques

A method includes obtaining at least one image and a ground truth map associated with the at least one image. The method also includes generating multiple optical maps using multiple algorithms and the at least one image. The method further includes, for each algorithm, identifying at least one score for the algorithm using one or more of the optical maps generated using the algorithm and the ground truth map. The ground truth map identifies one or more boundaries associated with one or more foreground objects in the at least one image. The scores identify how well the optical maps generated using the algorithms separate the one or more foreground objects from a background in the at least one image.

Systems and methods for quantitative evaluation of optical map quality and for data augmentation automation

A method includes processing, using at least one processor of an electronic device, multiple reference images of a scene using a first convolutional neural network (CNN) to generate a confidence map and a disparity map. The method also includes generating, using the at least one processor, an initial Bokeh image based on the disparity map and the reference images using a depth-of-field (DoF) renderer. The method further includes refining, using the at least one processor, the initial Bokeh image using a second CNN to generate a refined Bokeh image, where the second CNN uses the confidence map, the disparity map, and the reference images to generate the refined Bokeh image.

System and method for generating bokeh image for DSLR quality depth-of-field rendering and refinement and training method for the same

A method includes obtaining at least three input image frames of a scene captured using at least three imaging sensors. The input image frames include a reference image frame and multiple non-reference image frames. The method also includes generating multiple disparity maps using the input image frames. Each disparity map is associated with the reference image frame and a different non-reference image frame. The method further includes generating multiple confidence maps using the input image frames. Each confidence map identifies weights associated with one of the disparity maps. In addition, the method includes generating a depth map of the scene using the disparity maps and the confidence maps. The imaging sensors are arranged to define multiple baseline directions, where each baseline direction extends between the imaging sensor used to capture the reference image frame and the imaging sensor used to capture a different non-reference image frame.

Array-based depth estimation

A method includes obtaining, using at least one processor of an electronic device, multiple calibration parameters associated with multiple sensors of a selected mobile device. The method also includes obtaining, using the at least one processor, an identification of multiple imaging tasks. The method further includes obtaining, using the at least one processor, multiple synthetically-generated scene images. In addition, the method includes generating, using the at least one processor, multiple training images and corresponding meta information based on the calibration parameters, the identification of the imaging tasks, and the scene images. The training images and corresponding meta information are generated concurrently, different ones of the training images correspond to different ones of the sensors, and different pieces of the meta information correspond to different ones of the imaging tasks.

Multi-sensor, multi-view, multi-frame, multi-task synthetic image fusion engine for mobile imaging system

A method includes processing, using at least one processor of an electronic device, each of multiple images using a photometric augmentation engine, where the photometric augmentation engine performs one or more photometric augmentation operations. The method also includes applying, using the at least one processor, multiple layers of a convolutional neural network to each of the images, where each layer generates a corresponding feature map. The method further includes processing, using the at least one processor, at least one of the feature maps using at least one feature augmentation engine between consecutive layers of the multiple layers, where the at least one feature augmentation engine performs one or more feature augmentation operations.

Mobile data augmentation engine for personalized on-device deep learning system

A method includes identifying, by at least one processor, multiple features of input data using a common feature extractor. The method also includes processing, by the at least one processor, at least some identified features using each of multiple pre-processing branches. Each pre-processing branch includes a first set of neural network layers and generates initial outputs associated with a different one of multiple data processing tasks. The method further includes combining, by the at least one processor, at least two initial outputs from at least two pre-processing branches to produce combined initial outputs. In addition, the method includes processing, by the at least one processor, at least some initial outputs or at least some combined initial outputs using each of multiple post-processing branches. Each post-processing branch includes a second set of neural network layers and generates final outputs associated with a different one of the multiple data processing tasks.

Multi-task fusion neural network architecture

A method includes obtaining, using at least one processor, an input image frame. The method also includes identifying, using the at least one processor, one or more regions of the input image frame containing redundant information. In addition, the method includes performing, using the at least one processor, an image processing task using the input image frame. The image processing task is guided based on the one or more identified regions of the input image frame. The method may further include obtaining, using the at least one processor, a coarse depth map associated with the input image frame. Performing the image processing task may include refining the coarse depth map to produce a refined depth map, where the refining of the coarse depth map is guided based on the one or more identified regions of the input image frame.

Guided backpropagation-gradient updating for image processing task using redundant information from image

A method includes obtaining, using at least one processor, first and second input image frames, where the first and second input image frames are associated with first and second image planes, respectively. The method also includes obtaining, using the at least one processor, a depth map associated with the first input image frame. The method further includes producing another version of the depth map by performing one or more times: (a) projecting, using the at least one processor, the first input image frame to the second image plane in order to produce a projected image frame using (i) the depth map and (ii) information identifying a conversion from the first image plane to the second image plane and (b) adjusting, using the at least one processor, at least one of the depth map and the information identifying the conversion from the first image plane to the second image plane.

Super-resolution depth map generation for multi-camera or other environments

A method includes obtaining, using at least one processor of an electronic device, multiple video frames of a video stream and multiple depth frames corresponding to the multiple video frames. The method also includes generating, using the at least one processor, multiple blur kernel maps based on the multiple depth frames. The method further includes reducing, using the at least one processor, depth errors in each of the multiple blur kernel maps. The method also includes performing, using the at least one processor, temporal smoothing on the multiple blur kernel maps to suppress temporal artifacts between different ones of the multiple blur kernel maps. In addition, the method includes generating, using the at least one processor, blur effects in the video stream using the multiple blur kernel maps.

System and method for synthetic depth-of-field effect rendering for videos

A method includes obtaining, using first and second image sensors of an electronic device, first and second images, respectively, of a scene. The method also includes obtaining, using an image depth sensor of the electronic device, a third image and a first depth map of the scene, the first depth map having a resolution lower than a resolution of the first and second images. The method further includes undistorting the first and second images using the third image and the first depth map. The method also includes rectifying the first and second images using the third image and the first depth map. The method further includes generating a disparity map using the first and second images that have been undistorted and rectified. In addition, the method includes generating a second depth map using the disparity map and the first depth map, where the second depth map has a resolution that is higher than the resolution of the first depth map.

Apparatus and method for dynamic multi-camera rectification using depth camera

A method includes obtaining a first image of a scene using a first image sensor of an electronic device and a second image of the scene using a second image sensor of the electronic device. The method also includes generating a first feature map from the first image and a second feature map from the second image. The method further includes generating a third feature map based on the first feature map, the second feature map, and an asymmetric search window. The method additionally includes generating a depth map by restoring spatial resolution to the third feature map.

Growing community of inventors

Allen, TX, United States of America

Yingmao Li