Conversion and encoding techniques

• US 7,489,812 B2
• Filed: 06/07/2002
• Issued: 02/10/2009
• Est. Priority Date: 06/07/2002
• Status: Active Grant

• Current Assignee: HomeWAV, LLC
• Original Assignee: dynamic digital depth Research Pty Ltd
• Inventors: Harman, Philip Victor, Flack, Julien Charles, Fox, Simon Richard
• Primary Examiner(s): Lu, Tom Y
• Application Number:
  US10/164,005
• Publication Number:
  US 20030228034A1
• Time in Patent Office: 2,440 Days
• Field of Search: 382/154, 382/285, 348/43
• US Class Current: 382/154
• CPC Class Codes: G06T 7/55 from multiple images

Abstract

A method of creating at least one depth map for an image sequence including the steps of receiving depth data for a plurality of points in the image sequence, utilising the depth data and a classifier to ascertain depth characteristics as a function of image characteristics and relative position and creating a depth map for at least one frame of the image sequence utilising the image characteristics.

Claims

1. A method of creating at least one depth map for an image sequence including the steps of:

receiving depth data for a plurality of points across frames in said image sequence;

utilizing said depth data and a classifier to ascertain depth characteristics as a function of image characteristics including RGB values;

creating a depth map for at least one frame of said image sequence utilizing said image characteristics including the RGB values.

2. A method as claimed in claim 1, wherein said image characteristics further include relative positions.

3. A method as claimed in claim 1, further including an initial step of ascertaining depth data for at least a predetermined number of points within said image sequence.

4. A method as claimed in claim 1, including an initial step of bootstrapping said classifier.

5. A method of creating at least one depth map for an image sequence including the steps of:

dividing said image sequence into a plurality of image shots;

receiving depth data for a plurality of points in at least one of said image shots;

utilizing said depth data and a classifier to ascertain depth characteristics as a function of image characteristics and relative position;

creating a depth map for at least one frame of said at least one of said image shots utilizing said image characteristics.

6. A method as claimed in claim 5, wherein each shot is assigned a corresponding classifier.

7. A method as claimed in claim 6, wherein at least one corresponding classifier is initialized to a similar state as a previous classifier.

8. A method of creating at least one depth map for an image sequence including the steps of:

dividing said image sequence into a plurality of image shots;

for each image shot receiving depth data for a plurality of points and utilizing said depth data and a respective classifier to create an initial algorithm to ascertain depth characteristics as a function of image characteristics and relative position;

creating at least one combined algorithm by combining at least two initial algorithms from a plurality of said image shots;

creating a depth map for at least one frame of said image sequence using said at least one combined algorithm.

9. A method as claimed in claim 8, wherein said image characteristics include RGB values.

10. A method as claimed in claim 8, wherein said image characteristics include relative xy positions.

11. A method as claimed in claim 8, further including an initial step of ascertaining depth data for at least a predetermined number of points within said image sequence.

12. A method as claimed in claim 8, wherein said combined algorithm is determined using an average of said at least two initial algorithms.

13. A method as claimed in claim 8, wherein said combined algorithm is determined using a weighted average of said at least two initial algorithms.

14. A method as claimed in claim 8, including an initial step of initializing said classifier to a random state.

15. A method as claimed in claim 8, wherein at least one said respective classifier is initialized to a similar state as a previous classifier.