Object detection in images

• US 20070036432A1
• Filed: 11/08/2004
• Published: 02/15/2007
• Est. Priority Date: 11/12/2003
• Status: Active Grant

• Current Assignee: British Telecommunications PLC (BT Group PLC)
• Original Assignee: British Telecommunications PLC (BT Group PLC)
• Inventors: Xu, Li-Qun, Landabaso, Jose-Luis
• Granted Patent: US 7,742,650 B2
• Time in Patent Office: Days
• US Class Current: 382/173
• CPC Class Codes: G06T 2207/10016 Video; Image sequence
• CPC Class Codes: G06T 2207/20036 Morphological image processing
• CPC Class Codes: G06T 2207/30236 Traffic on road, railway or crossing
• CPC Class Codes: G06T 7/11 Region-based segmentation
• CPC Class Codes: G06T 7/194 involving foreground-background segmentation
• CPC Class Codes: G06V 10/273 removing elements interfering with the pattern to be recognised
• CPC Class Codes: G06V 10/28 Quantising the image, e.g. histogram thresholding for discrimination between background and foreground patterns
• CPC Class Codes: G06V 30/153 using recognition of characters or words

Abstract

The invention applies a segmentation operation to an input image to identify foreground objects of interest, and then applies a shadow removal operation to remove any detected shadows from the foreground segmentation. The shadow removal algorithms can leave holes and bisections in the segmentation map, however, which will then subsequently impact on an object detection step performed using connected component analysis. To get around this problem, the invention applies a conditional morphological dilation operation to the segmentation map to ‘grow’ the segmented blobs to fill in any holes and bisections, without re-growing shadow pixels in the segmentation. The result is an object detection method and system which is robust to illumination changes causing shadows and/or highlights.

Claims

1. An image processing method for detecting objects within an input image, the image being composed of picture elements, the method comprising:

a) segmenting picture elements representing a foreground object within the input image from those picture elements forming the image background using a first segmentation technique, wherein the picture elements segmented as foreground include elements representing a shadow or highlight of the object; and

b) segmenting picture elements which have the characteristics of a shadow or highlight of an object from those picture elements representing the foreground object using at least one other segmentation technique adapted to detect shadows and/or highlights;

the method being characterised by further comprising the steps of;

c) segmenting as foreground surrounding picture elements to those picture elements which are already segmented as foreground;

d) repeating step c) until picture elements which were not segmented as foreground after step a) would be or are segmented as foreground; and

then

e) detecting as objects groups of adjacent picture elements which have been segmented as foreground.

2. An image processing method as claimed in claim 1, and further characterised in that the segmenting step a) further comprises:

for each picture element of the input image which is to be segmented as foreground, validating the foreground segmentation by comparison of the picture element with probability models relating to surrounding picture elements, wherein the foreground segmentation is confirmed if at least half of the models indicate that the picture element is foreground.

3. A computer program or suite of computer programs arranged such that when executed by a computer they control the computer to perform the method of claim 1.

4. A computer readable storage medium storing a computer program or at least on of a suite of computer programs according to claim 3.

5. An image processing system for detecting objects within an input image, the image being composed of picture elements, the system comprising:

image processing means arranged to receive an input image to be processed, and to apply the following image processing operations thereto;

a) to segment picture elements representing a foreground or moving object within the input image from those picture elements forming the image background using a first segmentation technique adapted to detect differences in the input image from a general background image, wherein the picture elements segmented as foreground include elements representing a shadow or highlight of the object;

b) to segment picture elements which have the characteristics of a shadow or highlight of an object from those picture elements representing the foreground object using at least one other segmentation technique adapted to detect shadows and/or highlights;

and being characterised by further applying the following operations;

c) to repeatedly segment as foreground surrounding picture elements to those picture elements already segmented as foreground until picture elements which were not segmented as foreground after step a) would be or are segmented as foreground; and

then

e) to detect as objects groups of adjacent picture elements which have been segmented as foreground.

6. An image processing system as claimed in claim 5, and further characterised in that the image processing means is further arranged, for each picture element of the input image which is to be segmented as foreground as a consequence of the segmentation a), to validate the foreground segmentation by comparison of the picture element with probability models relating to surrounding picture elements, wherein the foreground segmentation is confirmed if at least half of the models indicate that the picture element is foreground.