SYSTEM AND METHOD OF REVISING DEPTH OF A 3D IMAGE PAIR

• US 20120293624A1
• Filed: 05/19/2011
• Published: 11/22/2012
• Est. Priority Date: 05/19/2011
• Status: Active Grant

• Current Assignee: National Taiwan University (Government of The Republic of China), Himax Technologies Ltd. (Himax Technologies Incorporated)
• Original Assignee: National Taiwan University (Government of The Republic of China), Himax Technologies Ltd. (Himax Technologies Incorporated)
• Inventors: CHEN, LIANG-GEE, WU, CHIEN, LI, CHUNG-TE, LAI, YEN-CHIEH, CHENG, CHAO-CHUNG, HUANG, LING-HSIU
• Granted Patent: US 8,629,901 B2
• Time in Patent Office: Days
• US Class Current: 348/46
• CPC Class Codes: H04N 13/128 Adjusting depth or disparity

Abstract

A method of revising depth of a three-dimensional (3D) image is disclosed. The method comprises the following steps: firstly, at least one initial depth map associated with one image of the 3D image pair based on stereo matching technique is received, wherein the one image comprises a plurality of pixels, and the initial depth map carries an initial depth value of each pixel. Then, the inconsistence among the pixels of the one image of the 3D image pair is detected to estimate a reliable map. Finally, the initial depth value is interpolated according to the reliable map and the proximate pixels, so as to generate a revised depth map by revising the initial depth value.

Claims

1. A system of revising depth of a three-dimensional (3D) image pair, comprising:

a depth generator configured to generate at least one initial depth map associated with one image of the 3D image pair based on stereo matching technique, wherein the one image comprises a plurality of pixels, and the initial depth map carries an initial depth value of each pixel; and

a depth revisor, comprising;

an inconsistence detection unit configured to detect the inconsistence among the pixels based on two-directional occlusion detection technique and estimate a reliable map according to the detected inconsistence; and

an interpolation unit configured to interpolate the initial depth value according to the reliable map and the proximate pixels, so as to generate a revised depth map by revising the initial depth value.

2. The system of claim 1, wherein the inconsistence detection unit computes a cost value for each of the pixels which denotes the inconsistence or discontinuity between one of the pixels and neighborhood nodes.

3. The system of claim 2, wherein the reliable map is composed of a group of reliable points and a group of unreliable points, the reliable points comprises the pixels, the cost value of which is lower than a pre-defined threshold, and the unreliable points comprises the pixels, the cost value of which is higher than the pre-defined threshold.

4. The system of claim 3, wherein the interpolation unit interpolates the initial depth value by spatial and luminance proximity of the pixels for the unreliable points.

5. The system of claim 3, wherein the reliable map carries a reliable value for every pixel, and the inconsistence detection unit assigns 1 to the reliable values for the reliable points and assigns 0 to the reliable values for the unreliable points.

6. The system of claim 1, wherein the depth generator generates a left depth map corresponding to a left image and generates a right depth map corresponding to a right image, wherein the one image of the 3D image pair is the left image or the right image.

7. The system of claim 1, further comprising a depth-image-based rendering (DIBR) unit configured to receive the revised depth map and accordingly generate at least one revised left image and at least one revised right image.

8. The system of claim 1, the inconsistence detection unit utilities Left-Right-checking or Right-Left-checking techniques to detect the inconsistence.

9. The system of claim 1, the interpolation unit comprises a trilateral filter.

10. A method of revising depth of a three-dimensional (3D) image pair, comprising:

receiving at least one initial depth map associated with one image of the 3D image pair based on stereo matching technique, wherein, the image comprises a plurality of pixels, and the initial depth map carries an initial depth value of each pixel;

detecting the inconsistence among the pixels of the one image of the 3D image pair to estimate a reliable map; and

interpolating the initial depth value according to the reliable map and the proximate pixels, so as to generate a revised depth map by revising the initial depth value.

11. The method of claim 10, wherein the step of estimating the reliable map comprises:

computing a cost value for each of the pixels which denotes the inconsistence or discontinuity between one of the pixels and neighborhood nodes; and

estimating the reliable map according to the detected inconsistence.

12. The method of claim 11, wherein the reliable map carries a reliable value for every pixel, and the step of estimating the reliable map further comprises:

providing a pre-defined threshold;

assigns 1 to the reliable values for the pixels, the cost value of which is lower than the pre-defined threshold; and

assigns 0 to the reliable values for the pixels, the cost value of which is higher than the pre-defined threshold.

13. The method of claim 12, wherein the step of interpolating the initial depth value comprises:

computing the initial depth value by spatial and luminance proximity of the pixels.

14. The method of claim 10, implemented by Left-Right-checking or Right-Left-checking techniques to detect the inconsistence.

15. The method of claim 10, utilizing a trilateral filter to interpolate the initial depth value.

16. The method of claim 10, wherein said at least one initial depth map comprises a left depth map corresponding to a left image and a right depth map corresponding to a right image, wherein the one image of the 3D image pair is the left image or the right image.

17. The method of claim 10, further comprising:

receiving the revised, depth map and accordingly generating at least one revised left image and at least one revised right image.