3D system including a neural network

• US 10,277,877 B2
• Filed: 10/14/2016
• Issued: 04/30/2019
• Est. Priority Date: 11/13/2015
• Status: Active Grant

• Current Assignee: VEFXi Corporation
• Original Assignee: VEFXi Corporation
• Inventors: Peterson, Craig, Doyle, Pat, Roberts, Markus, Lomov, Sergey
• Primary Examiner(s): Young, Patricia I
• Application Number:
  US15/293,382
• Publication Number:
  US 20170142394A1
• Time in Patent Office: 928 Days
• Field of Search: 348 43
• CPC Class Codes: H04N 13/261 with monoscopic-to-stereoscopic image conversion

Abstract

A three dimensional system that includes a neural network.

Claims

1. A method for conversion of a series of two dimensional images using a processor into a series of three dimensional images comprising:

(a) said processor receiving said series of two dimensional images where each of said series of two dimensional images are from a same viewpoint, where said series of two dimensional images do not include any images from a viewpoint other than said same viewpoint;

(b) said processor processing said series of two dimensional images all of which are from said same viewpoint and do not include any images from a viewpoint other than said same viewpoint based upon a neural network to determine a respective depth map associated with each of said series of two dimensional images where all of said series of two dimensional images are from said same viewpoint, where said series of two dimensional images do not include any images from a viewpoint other than said same viewpoint;

(c) said processor processing said depth map to render said two dimensional images as said series of three dimensional images for being displayed on a 3D display;

(d) rendering said two dimensional images all of which are from the same viewpoint and do not include any images from a viewpoint other than said same viewpoint based upon said depth map on a display;

(e) wherein said neural network includes a plurality of interconnected digital neurons which mimic a logical operation of biological brain neurons with synapse connections with other brain cells for a purpose of video analysis and decision making;

(f) wherein said neural network includes updating said interconnected neuron synapse using at least one of genetic, back-propagation, and adaptive learning techniques;

(g) wherein said neural network includes a set of adaptive weights, thresholds, and activation functions;

(h) wherein said neural network receives image color space components of said series of two dimensional images; and

(i) wherein said image color space components are in a substantially intensity independent color space.

2. The method of claim 1 wherein portions of said series of three dimensional images has an appearance of being behind a plane of a screen of said display and other portions of said series of three dimensional images has the appearance of being in front said plane of said screen of said display.

3. The method of claim 1 wherein said adaptive weights, thresholds, and activation functions approximate non-linear functions of inputs to said neural network.

4. The method of claim 3 wherein said neural network includes at least one of genetic training, back propagation, and genetic learning modes.

5. The method of claim 3 wherein said neural network includes forward propagation.

6. The method of claim 3 wherein said neural network includes recurrent structures.

7. The method of claim 3 wherein said neural network includes back propagation, forward propagation, and recurrent structures.

8. The method of claim 1 wherein said neural network receives pixel values of said series of two dimensional images.

9. The method of claim 1 wherein said neural network receives texture information of said series of two dimensional images.

10. The method of claim 1 wherein said neural network receives edge characteristics of said series of two dimensional images.

11. The method of claim 1 wherein said neural network receives structure information of said series of two dimensional images.

12. The method of claim 1 wherein said neural network receives facial feature information of said series of two dimensional images.