Foreground detection using intrinsic images

• US 20060126933A1
• Filed: 12/15/2004
• Published: 06/15/2006
• Est. Priority Date: 12/15/2004
• Status: Active Grant

• Current Assignee: Mitsubishi Electric Research Laboratories, Inc. (Mitsubishi Electric Corporation)
• Original Assignee: Mitsubishi Electric Research Laboratories, Inc. (Mitsubishi Electric Corporation)
• Inventors: Porikli, Fatih M.
• Granted Patent: US 7,359,552 B2
• Time in Patent Office: Days
• US Class Current: 382/173
• CPC Class Codes: G06T 7/194 involving foreground-background segmentation
• CPC Class Codes: G06T 7/215 Motion-based segmentation
• CPC Class Codes: G06V 10/28 Quantising the image, e.g. histogram thresholding for discrimination between background and foreground patterns

Abstract

A temporal sequence of images is acquired of a dynamic scene. Spatial gradients are determined using filters. By taking advantage of the sparseness of outputs of the filters, an intrinsic background image is generated as median filtered gradients. The intrinsic background image is then divided into the original sequence of images to yield intrinsic foreground images. The intrinsic foreground images can be thresholded to obtain a detection mask.

Claims

1. A method for modeling background and foreground regions of a scene from a sequence of input images of the scene, comprising:

generating an intrinsic background image from a sequence of input images acquired of a scene; and

dividing each input image by the intrinsic background image to obtain a sequence of intrinsic foreground images.

2. The method of claim 1, further comprising:

updating the intrinsic background image as each input image is processed.

3. The method of claim 1, in which a sampling period of the input images varies over time.

4. The method of claim 1, in which a sampling period of the input images varies as a function of moving objects in the scene.

5. The method of claim 1, further comprising:

converting each input image to a logarithmic image.

applying a filter to each logarithmic image;

determining a median image from a temporal window of filtered logarithmic images;

reverse filtering the median image;

subtracting the median image from a current logarithmic image to obtain a logarithmic foreground image;

applying an inverse logarithm to the logarithmic foreground image to obtain the intrinsic foreground image; and

applying the inverse logarithm to the median image to obtain the intrinsic background image.

6. The method of claim 5, in which the filter is a spatial derivative filters.

7. The method of claim 5, in which the filter is a line detector.

8. The method of claim 1, further comprising:

detecting moving objects in the intrinsic foreground image; and

generating a mask image for the moving objects.

9. The method of claim 8, in which the mask image is based on intensity statistics of a difference between the intrinsic background image and the intrinsic foreground images.

10. The method of claim 1, further comprising:

detecting lines and silhouettes from the intrinsic background image using a line detector.