Estimating scenes using statistical properties of images and scenes

• US 6,263,103 B1
• Filed: 11/30/1998
• Issued: 07/17/2001
• Est. Priority Date: 11/30/1998
• Status: Expired due to Fees

• Current Assignee: Mitsubishi Electric Research Laboratories, Inc. (Mitsubishi Electric Corporation)
• Original Assignee: Mitsubishi Electric Research Laboratories, Inc. (Mitsubishi Electric Corporation)
• Inventors: Freeman, William T., Pasztor, Egon C.
• Primary Examiner(s): Patel, Jayanti K.
• Application Number:
  US09/203,108
• Time in Patent Office: 960 Days
• Field of Search: 382/100, 382/156, 382/158, 382/160, 382/171-173, 382/195, 382/228, 382/284, 382/159, 382/181, 382/190, 382/294, 382/103, 342/159, 342/25, 342/64, 348/700, 345/433-435
• US Class Current: 382/173
• CPC Class Codes: G06T 7/41 based on statistical description of texture
• CPC Class Codes: G06V 20/00 Scenes; Scene-specific elements control of digital cameras H04N23/60

Abstract

A method for estimating scenes from images generates a plurality of scenes and renders an image for each scene. The scenes and corresponding images are partitioned into patches. Each patch is quantitized as a vector and probability density function is fitted to each vector. The patches and probability densities are organized as a Markov network where local probability information is propagated to neighboring nodes. After propagation, the probability density at each node is used to estimate the scene.

Claims

1. A method for estimating scenes from images, comprising the steps of:

generating a plurality of scenes and rendering an image for each scene;

partitioning each scene and corresponding image into patches;

quantifying each patch as a vector and modeling a probability density to each vector;

representing the patches and probability densities as a network;

iteratively propagating local probability information to neighboring nodes of the network; and

reading the probability density at each node to estimate the scene.

2. The method of claim 1 wherein the scenes and images are synthetically generated.

3. The method of claim 2 wherein the scenes and images are generated by computer graphics.

4. The method of claim 1 wherein the scenes and images are randomly generated.

5. The method of claim 1 wherein the patches can have multiple sizes.

6. The method of claim 1 wherein the patches are redundantly overlaid.

7. The method of claim 1 wherein the patches are organized as a gaussian pyramid having a plurality of resolution levels.

8. The method of claim 1 wherein the vectors are determined by principle components analysis.

9. The method of claim 1 wherein the network is a Markov chain.

10. The method of claim 9 wherein nodes of the network represent the patches and the associated probability densities, and edges connecting the nodes represent statistical dependencies between the nodes.

11. The method of claim 10 wherein local probabilities are propagated by factorizing joint probabilities of neighboring nodes of the network.