Mitsubishi Electric Research Laboratories, Inc.

Mitsubishi Electric Corporation

Suhas Lohit

Petros T Boufounos

Hassan Mansour

Dehong Liu

Michael Jones

Kuan-Chuan Peng

Pedro Miraldo

Moitreya Chatterjee

Embodiments for rendering a 2D image of a dynamic 3D scene from different view angles using a neural radiance field (NeRF) are provided. In this regard, an AI image processing system is configured to process coordinates of a point in a dynamic 3D scene with an RNN over a number of time steps indicated by a time instance of interest to produce motion information of the point at the time instance of interest, process the motion information with a fully connected neural network to produce a displacement of the point from the coordinates in the dynamic 3D scene, and process a displaced point from a view angle of interest with the NeRF to render the point on the 2D image of the dynamic 3D scene at the time instance of interest. The displaced point is generated based on an estimate of the displacement of the point by leveraging motion cues.

Rendering two-dimensional image of a dynamic three-dimensional scene

The present disclosure provides a system and a method for generating a radar image of a scene. The method comprises receiving radar measurements of a scene collected from a set of antennas, wherein the set of antennas are under uncertainties caused by one or a combination of position ambiguities and clock ambiguities of each of the antennas. The method further comprises generating the radar image of the scene by solving a sparse recovery problem. The sparse recovery problem determines, until a termination condition is met, a set of image shifts of the radar image corresponding to different uncertainties of the antennas and updates an estimate of the radar image, based on the determined set of image shifts of the radar image. The sparse recovery problem is solved with a neural network denoiser that denoises a filtering of the estimate of the radar image.

System and method for generating a radar image of a scene

Systems, methods and apparatus for image processing for reconstructing a super resolution (SR) image from multispectral (MS) images. A processor to iteratively, fuse a MS image with an associated PAN image of the scene. Each iteration includes using a gradient descent (GD) approach with a learned forward operator, to generate an intermediate high-resolution multispectral (IHRMS) image with an increased spatial resolution and a smaller error to the DSRMS image compared to the stored MS image. Project the IHRMS image using a trained convolutional neural network (CNN) to obtain an estimated synthesized high-resolution multispectral (ESHRMS) image, for a first iteration. Use the ESHRMS image and the PAN image, as an input to the GD approach for following iterations. The updated IHRMS image is an input to another trained CNN for the following iterations. After predetermined number of iterations, output the fused high-spatial and high-spectral resolution MS image.

Systems and methods for multi-spectral image fusion using unrolled projected gradient descent and convolutinoal neural network

Growing community of inventors

Tempe, AZ, United States of America

Suhas Lohit