Intel Corporation

Fan Zhang

Oscar Nestares

Bindita Chaudhuri

Xin Zeng

Joshua J Ratcliff

Namakkal N Venkatesan

Kalpana Seshadrinathan

Cunming Liang

Bruce Richardson

Maha El Choubassi

Roger Keith Wiles

Vladan Popovic

Techniques related to calibrating fisheye cameras using a single image are discussed. Such techniques include applying a first pretrained convolutional neural network to an input fisheye image to generate camera model parameters excluding a principle point and applying a second pretrained convolutional neural network to the fisheye image and a difference of the fisheye image and a projection of the fisheye image using the camera model parameters to generate the principle point.

Single image ultra-wide fisheye camera calibration via deep learning

An example system for generating light field content is described herein. The system includes a receiver to receive a plurality of images and a calibrator to intrinsically calibrate a camera. The system also includes a corrector and projector undistort the images and project the undistorted images to generate undistorted rectilinear images. An extrinsic calibrator may rectify and align the undistorted rectilinear images. Finally, the system includes a view interpolator to perform intermediate view interpolation on the rectified and aligned undistorted rectilinear images.

Generating virtual reality content via light fields

An example system for generating stereoscopic light field panoramas includes at least one memory; and at least one processor to execute instructions to: access images of a scene captured by a plurality of cameras, the images including light field information associated with the scene; generate, based on first portions of the images, a first stereoscopic light field panorama for a first perspective associated with a left eye viewpoint of the scene, the left eye viewpoint associated with a left eye position on a first circle; and generate, based on second portions of the images, a second stereoscopic light field panorama for a second perspective associated with a right eye viewpoint of the scene, the right eye viewpoint associated with a right eye position on a second circle, the first circle being different than and concentric to the second circle.

Generating stereoscopic light field panoramas using concentric viewing circles

There is disclosed in one example a computing apparatus, including: a hardware platform configured to communicatively couple with a multi-tenant cloud service, the multi-tenant cloud service including an oversubscribable resource; and a service assurance for oversubscribable resource (SAOR) engine configured to: receive tenant subscriptions to the oversubscribable resource, wherein tenant subscriptions exceed available instances of the oversubscribable resource; receive per-tenant quality of service (QoS) metrics for the oversubscribable resource; receive an allocation request from a guest for allocation of an instance of the oversubscribable resource; compare the request to currently-available instances of the oversubscribable resource; determine that the oversubscribable resource has capacity to service the request according to the QoS metrics of the tenant; and allocate an instance of the oversubscribable resource to the guest.

Oversubscribable resource allocation

Techniques related to generating a virtual view of a scene from a position between positions of known input images for presentation to a viewer are discussed. Such techniques include applying a convolution mask that approximates an inverse of a linear combination of at least the horizontal and vertical convolution matrices representative of gradient detection in the input images to a virtual intermediate gradient image that is a combination of at least an interpolated virtual image at the position, an interpolated horizontal gradient map at the position, and an interpolated vertical gradient map at the position to generate a final virtual image for the position.

Virtual view interpolation between camera views for immersive visual experience

Techniques related to generating a virtual view of a scene from a position between positions of known input images for presentation to a viewer are discussed. Such techniques include applying gradient-based hole-filling interpolation.

Method and system of robust virtual view generation between camera views

An example system for generating stereoscopic light field panoramas includes a receiver to receive a plurality of synchronized images. The system also includes a calibrator and projector to calibrate the synchronized images, undistort the synchronized images, and project the undistorted images to a sphere to generate undistorted rectilinear images. The system further includes a disparity estimator to estimate a disparity between neighboring views of the undistorted rectilinear images to determine an optical flow between the undistorted rectilinear images. The system includes a view interpolator to perform in-between view interpolation on the undistorted rectilinear images based on the optical flow. The system also further includes a light field panorama generator to generate a stereoscopic light field panorama for a plurality of perspectives using concentric viewing circles.

Examples include a processor including fetch circuitry to fetch a guest physical address translation instruction having a format with fields to specify at least an opcode and locations of a source vector and a destination vector, decode circuitry to decode the fetched guest physical address translation instruction, and execution circuitry to execute the decoded guest physical address translation instruction. Execution of the decoded guest physical address translation instruction includes comparing guest physical addresses of the source vector with base and end addresses of a selected memory region, masking a guest physical address of the source vector if the guest physical address is in the selected memory region, translating the masked guest physical addresses into host addresses, and storing the host addresses into the destination vector.

Processing vectorized guest physical address translation instructions

Techniques related to interpolating an intermediate view image from multi-view images are discussed. Such techniques include downsampling first and second images that represent a view of a scene, generating a disparity map based on applying a first CNN to the downscaled first and second images, translating the downscaled first and second images using the disparity map, applying a second CNN to the translated downscaled first and second images and the disparity map to generate a downscaled intermediate image, and upscaling the downscaled intermediate image to an intermediate image at the resolution of the first and second images using an image super-resolution convolutional neural network.

View interpolation of multi-camera array images with flow estimation and image super resolution using deep learning

Techniques related to image rectification for fisheye images are discussed. Such techniques may include iteratively determining warpings for equirectangular images corresponding to the fisheye images using alternating feature mappings between neighboring ones of the equirectangular images until mean vertical disparities between the features are reduced below a threshold, and warping the equirectangular images to rectified equirectangular images.

Circular fisheye camera array rectification

Techniques related to multiview video stabilization are discussed. Such techniques may include performing single view video stabilization on a reference video stream and propagating the single view video stabilization to another video stream based on a target disparity map.

Growing community of inventors

San Mateo, CA, United States of America

Fan Zhang