US 2003/0194110 A1 · Discriminating between changes in lighting and movement of objects in a series of images using different methods depending on optically detectable surface characteristics

Abstract

In a motion detection system, filters are applied to a series of digital images in order to determine whether changes in the properties of pixels between the current image and a reference model are due to motion of objects or to changes in lighting. The filter may be based on scaled differences between a current image and a reference model. The type of surface may be determined based on optical properties of the surface and then only the best filter for that type of surface is applied. Alternately, multiple filters may all be applied and the results combined. The processing may be reduced using a background model to determine which pixels are in the background and which are in the foreground and then only applying the filter to the foreground pixels.

Drawings

Claims27

Dependent claims

A method comprising the steps of:

from a series of multiple images of multiple pixels, selecting a current image X;
determining a reference image R of multiple pixels;
for each pixel j of multiple pixels of the current image X, determining a scaled difference SDIFFj, where
10 SDIFF j ≈ 1 M ⁢ ∑ i = 1 M ⁢ ⁢ ( x i + ( r _ - x _ ) - r i ) 2 / D i 2 , x _ = 1 N ⁢ ∑ i = 1 N ⁢ x i , r _ = 1 O ⁢ ∑ i = 1 O ⁢ r i ,
xi is the value of a property of pixel i of a region of M pixels around pixel j in the current image X, ri is the value of the property of a corresponding pixel i around a corresponding pixel j in the reference image R, Di is a noise factor depending on the sum of multiple differences between the value of the property of pixel i in different images or the reference image.

2 more elements

A method, comprising the steps of:

selecting a current image X of multiple pixels;
determining a reference image R of multiple pixels;
for each pixel j of multiple pixels of the current image X:
classifying a region of pixels around the pixel j in the current image X, into one of at least a first type and a second type of region, depending on the uniformity of a property of the pixels among the pixels in the region of pixels;
when the region is of the first type, first determining whether change in the property of pixel j indicates a change in lighting or movement of an object depending on a first function;
and
when the region is of the second type, second determining whether the change in the property of pixel j indicates a change in lighting or movement of an object depending on a second function, the second determining is different than the first determining, the difference including that the second function is different than the first function.

4 more elements

A method, comprising the steps of:

selecting a current image X of multiple pixels;
determining a reference image R of multiple pixels;
for each pixel j of multiple pixels of the current image:
first determining whether a change in the property of the pixel j indicates a change in lighting depending on a first function;
and
second determining whether the change in the property of the pixel j may indicates a change in lighting, depending on a second function, the second determining is different than the first determining, the difference including that the second function is different than the first function;
determining that the change in the property of the pixel j indicates a change in lighting if either the first determining or the second determining indicate that the change in the property of the pixel j indicates a change in lighting;
and otherwise that the change in the property of pixel j indicates the movement of an object.

5 more elements

A method, comprising the steps of:

from a series of multiple images, selecting a current image;
determining a background model, dependent on a plurality of the multiple images, for determining whether multiple pixels of the current image are in a background that has a low probability of moving or in a foreground that has a high probability of moving;
the background model depending on a probability function based on a Gaussian function;
determining which pixels of multiple pixels in the current image are in the foreground depending on the background model and the values of the properties of pixels of the current image;
determining a reference model, dependent on a plurality of the multiple images, for determining whether the pixels in the foreground of the multiple pixels of the current model, the reference model being dependent on an exponentially decaying average of previous frames
determining whether changes in properties of pixels that were determined to be in the foreground are due to changes in lighting or movement of objects depending on the reference model.

3 more elements

Dependencies are inferred from the claim text.

Description4,937 words

Summary853 words

#The present invention relates to the field of video motion detection and object tracking systems.

#Video motion detectors have been proposed for detecting moving objects based on changes between subsequent images. More specifically typical proposed systems incorporate a background model and identify objects based on changes between a current image and the background model. However, in addition to true moving objects, many changes between subsequent images are caused by changes in lighting. These include, for example, turning lights on or off, passing clouds casting shadows, moving lights such as car headlights shining on objects or background, and the changing position of the sun. The appearance of background objects can change dramatically due to slight changes in orientation with respect to a source of light. For example slight rotations of leaves of plants with respect to a light source or slight movements of highly reflective surfaces such as waves on water. In addition, light changes may be due to automatic adjustments of a video camera such as adjustment of the iris or adjustment of the gain of the video signal.

#Systems have been proposed that determine whether changes between images are caused by a change in lighting or the motion of objects based on normalized cross correlation NCC between the background model and the current image.

Brief description of the drawings7 figures

#These and other objects and advantages of the present invention will become clear to those skilled in the art in view of the detailed description of the following drawings:

#FIG. 1 shows a discriminator of the invention which uses the scaled difference to differentiate changes in lighting from movement of objects.

#FIG. 2 shows a discriminator of the invention in which a classifier selects the best of multiple filters to differentiate changes in lighting from movement of objects.

Detailed description3,888 words

#The following detailed description presents several example embodiments of the invention in order to describe the best mode for carrying out the invention and to enable those skilled in the art to make and use the invention.

#The same numbers have been used in different figures to indicate that components to which they refer in different figures may be similar so that previous descriptions of the components apply.

#In FIG. 1, a sequence of digital video images Ih of multiple pixels arrives at input 100. A current image X is selected and stored in memory 102 for processing. Preferably each subsequent image is processed so that the selected image is the last image that has arrived. The images also arrive at a reference model which is updated. The reference model unit 102 may provide a reference image R with the same number of pixels as the current image. Preferably the reference model is an exponentially decaying average of the previous images. For each image Ih where h=1 to k images, a reference image Jh is determined. The current image X is the last image that was received Ik. The current reference image R for use with the current image X being Jk−1, where 1 J h = α * J h - 1 + ( 1 - α ) ⁢ I h ⁢ [LeftBracketingBar] h = 1 k ,

Facts

Application

10/123,330Filed 16 Apr 2002

Published

16 Oct 2003

Term ends

Not estimatedThe source records don’t support an estimate.

Classification5 codes

G06T 7/254
Analysis of motion: involving subtraction of images
G06T 7/277
Analysis of motion: involving stochastic approaches, e.g. using Kalman filters
G06V 10/28
Image preprocessing: quantising the image, e.g. histogram thresholding for discrimination between background and foreground patterns
G06T
Image data processing or generation, in general

Timeline

16 Apr 2002Application filedUS 10/123,330
16 Apr 2002Assigned to Koninklijke Philips Electronics N.V.assignment of assignor's interest
16 Oct 2003Published
4 Feb 2009Assigned to Ipg Electronics 503 Limitedassignment of assignor's interest
20 Jul 2012Assigned to Pendragon Wireless LLCassignment of assignor's interest

Discriminating between changes in lighting and movement of objects in a series of images using different methods depending on optically detectable surface characteristics

Life of this patent

Discriminating between changes in lighting and movement of objects in a series of images using different methods depending on optically detectable surface characteristics

Life of this patent

Life of this patent

Abstract

Drawings

Drawings

Claims27

Description4,937 words

Drawings

Classification5 codes

Life of this patent

Description4,937 words