Detecting silhouette edges in images

• US 20040184677A1
• Filed: 03/19/2003
• Published: 09/23/2004
• Est. Priority Date: 03/19/2003
• Status: Active Grant

• Current Assignee: Mitsubishi Electric Research Laboratories, Inc. (Mitsubishi Electric Corporation)
• Original Assignee: Mitsubishi Electric Research Laboratories, Inc. (Mitsubishi Electric Corporation)
• Inventors: Raskar, Ramesh, Yu, Jingyi
• Granted Patent: US 7,206,449 B2
• Time in Patent Office: Days
• US Class Current: 382/284
• CPC Class Codes: G06T 11/00 2D [Two Dimensional] image generation
• CPC Class Codes: G06T 2207/10152 Varying illumination
• CPC Class Codes: G06T 2207/20192 Edge enhancement; Edge preservation
• CPC Class Codes: G06T 5/50 using two or more images, e.g. averaging or subtraction

Abstract

A method detects silhouette edges in images. An ambient image is acquired of a scene with ambient light. A set of illuminated images is also acquired of the scene. Each illuminated image is acquired with a different light source illuminating the scene. The ambient image is combined with the set of illuminated to detect cast shadows, and silhouette edge pixels are located from the cast shadows.

Claims

1. A method for detecting silhouette edges in images, comprising:

acquiring an ambient image of a scene with ambient light;

acquiring a set of illuminated images of the scene, each illuminated image acquired with a different light source illuminating the scene;

combining the ambient image with the set of illuminated to detect cast shadows; and

locating silhouette edge pixels from the cast shadows.

2. The method of claim 1 further comprising:

subtracting the ambient image from each illuminated image to obtain a corresponding difference image;

generating a maximum image from the set of difference images;

dividing each difference image by the maximum image to obtain a corresponding ratio image;

identifying an epipole pixel in each ratio image;

traversing, for each ratio image, radially outwards from the epipole to identify pixels with negative intensity transition; and

marking the pixels with the negative intensity transition as a silhouette edge pixels.

3. The method of claim 1 wherein the point light sources are flash units distributed evenly around a center of projection of a lens of a camera used to acquire the set of set of images.

4. The method of claim 2 wherein the flash units are mounted on a front surface of the camera substantially close to the center of projection.

5. The method of claim 1 wherein the silhouette edge pixels are located at a depth discontinuity in the scene.

6. The method of claim 1 wherein the maximum image is an approximation of an image with a point light source at a center of projection of a lens of a camera used to acquire the images.

7. The method of claim 1 wherein the maximum image is an approximation of an image with a point light source at a center of projection of a lens of a camera used to acquire the images.5. The method of claim 1 wherein the ambient image is I_ambent, the set of illuminated images are I′

_k for k=1, . . . , n, the point light source are located at positions P_k, the difference images are I_k=I_k′

−

I_ambient, for pixels (x) the maximum image I_max(x)=max_k(I_k(x)), k=1, . . . , n, and the ratio images are for all pixels (x), R_k(x)=I_k(x)/I_max(x).

8. The method of claim 7 wherein the ratio (I_k/I_max) is substantially to 1 in areas of the scene illuminated by the point light sources, and substantially zero in shadowed areas.

9. The method of claim 1 further comprising:

applying a one-dimensional edge detector along an epipolar line from the epipolar pixel to a given pixel to identify the negative intensity transitions.

10. The method of claim 5 wherein a width of the attached shadow is d=f(z₂−

z₁)B/(z₁·

z₂), where f is the focal length, B is a baseline. and z₁ and z₂ are depths a shadowing and shadowed edge.

11. The method of claim 1 further comprising:

setting an intensity of each pixel in the maximum image corresponding to the silhouette edge pixels to one.