Image analysis – Image transformation or preprocessing – Combining image portions
Reexamination Certificate
1998-08-05
2003-02-04
Mehta, Bhavesh (Department: 2721)
Image analysis
Image transformation or preprocessing
Combining image portions
C345S006000, C345S419000, C348S042000, C382S154000
Reexamination Certificate
active
06516099
ABSTRACT:
The present invention relates to an image processing apparatus and method.
Creating three-dimensional computer models of a real-life object has traditionally been time consuming and expensive, requiring skilled personnel and/or expensive equipment. One reason for this is that it is necessary to determine the relative “depth” of points on the object as well as the relative positions of the points in two dimensions.
Known techniques for creating three-dimensional computer models from real objects fall into one of two categories, namely active techniques in which depth information is obtained by actively sensing the surface of the object, and passive techniques in which depth information is obtained from images of the object.
Examples of active techniques include scanning the object with a pulsed laser beam and measuring the detection time of pulses relative to their transmission time to determine depth information (as in a laser “rangefinder”), and touching the object at a number of points on its surface with a position-sensitive probe. In conventional passive techniques, at least two images of the object taken from different camera positions are needed. To construct a three-dimensional computer model from the images, it is necessary to know firstly the location in each image of points which represent the same actual point on the object, and secondly the relative positions from which the images were taken. These are particularly onerous requirements. As a result, in known passive systems, distinguishing marks/calibrations are added to the object or its surroundings to enable matching points to be easily identified in the images, and/or the images are taken from known camera positions.
For example, WO-A-90/10194 discloses a system for measuring strain distribution in an object using the three-dimensional coordinates of points on the object surface calculated from two images of the object. To facilitate the matching of points in the images, a uniform square grid pattern is applied to the object before the images are taken, by electrochemical etching or silk screening. Corresponding points of intersection of the grid lines can then be easily identified in the images. Further, the object is placed on a rotary table which is rotated by a known angle between images, thereby defining the relative camera positions.
U.S. Pat. No. 4,803,645 discloses a system for determining three-dimensional coordinates of points on an object in which a grid or similar periodic pattern is projected onto the object using a light projector, and images of the object are taken using three imaging systems at fixed, predefined positions.
WO-A-88/02518 discloses a system for producing a depth map of an object from a plurality of images taken from imaging devices set in a pre-defined, known configuration. Similarly, U.S. Pat. No. 5,307,136 discloses an automobile distance detection system which determines the range of an automobile using images taken from a plurality of cameras mounted in a pre-defined, known configuration in the user's automobile.
GB-A-2244621, WO-A-92/06444 and U.S. Pat. No. 4,695,156 all disclose systems for determining three-dimensional coordinates of points on an object surface from stereo images of the object taken at known camera positions.
In many cases, however, it is inconvenient, expensive, and/or infeasible to provide a reference grid or markings/calibrations on the object or its surroundings, or to take images from known relative positions. Reliable and accurate techniques are then required to match points in the images and/or to calculate the relative camera positions.
Even if matching points are identified using reference features added to the object or its surroundings, an accurate and reliable technique for determining the positions at which the images were taken is necessary if the images were not taken from known relative positions.
Whether or not reference markings on the object or its surroundings and/or known camera positions are used, when a passive technique is employed, a method which accurately generates three-dimensional coordinate points using the matching image points and the camera positions is needed. This is particularly important, however, when no reference markings are used and the camera positions are not known, since inaccuracies in the matching points identified in the images and inaccuracies in the calculated camera positions can cause inaccuracies in the calculated three-dimensional coordinates of points.
The present invention aims to address one or more of the above problems.
In one aspect, the present invention aims to provide an apparatus and method for determining the positions at which images of an object were taken.
In another aspect, the present invention aims to provide an apparatus and method for determining points in a three-dimensional space, using matching image features and the relative positions of the images.
The present invention provides an image processing apparatus or method in which the positional relationship between first, second and third images is determined a plurality of times using matching features in the images, each time on the basis of a different combination of transformations between the three images, with the most accurate relationship being selected. Preferably, the transformation combinations considered are affine-affine, affine-perspective, perspective-perspective and perspective-affine, or if the transformation between one pair of images is fixed, known transformation-affine and known transformation-perspective.
The present invention provides an image processing apparatus or method in which signals defining features matched using different techniques in first and second images are processed without using prior information on the relationship between the positions from which the images were taken to determine the relationship. The relationship is determined a plurality of times, each time using features matched with a different matching technique or techniques, to give a selection of relationships.
The present invention provides an image processing apparatus or method in which signals defining features matched in first and second images are processed without using prior information on the relationship between the positions from which the images were taken to determine the relationship. The following steps are performed a plurality of times: matched features are used to set up a non-physically realisable matrix (such as the fundamental matrix, for example), which is then converted into a physically realisable matrix (such as the physical fundamental matrix, for example) and its accuracy is tested. The most accurate physically realisable matrix is selected.
The present invention provides an image processing apparatus or method in which signals defining corresponding object features in at least three images of the object and signals defining the relationships between the imaging positions are processed to produce points in a three-dimensional space representing points on the object. Each pair of corresponding features is used to define a point in the 3D space, and the 3D points produced from the same corresponding features are used to define a 3D point representing a point on the object. Processing is then performed to see if the 3D object points could represent the same point on the object, and, if they could, further processing is performed.
REFERENCES:
patent: 3901595 (1975-08-01), Helava et al
patent: 4175862 (1979-11-01), DiMatteo et al.
patent: 4654872 (1987-03-01), Hisano et al.
patent: 4695156 (1987-09-01), Taft
patent: 4803645 (1989-02-01), Ohtomo et al.
patent: 4825393 (1989-04-01), Nishiya
patent: 4935810 (1990-06-01), Nonami et al.
patent: 5144373 (1992-09-01), Moler
patent: 5166878 (1992-11-01), Poelstra
patent: 5202928 (1993-04-01), Tomita et al.
patent: 5307136 (1994-04-01), Saneyoshi
patent: 5381526 (1995-01-01), Ellson
patent: 5390291 (1995-02-01), Ohashi
patent: 5454069 (1995-09-01), Knapp et al.
patent: 5519485 (1996-05-01), Ohtani et al.
patent: 5559334 (1996-09-01), Gupta et al.
patent: 5598515
Davison Allan Joseph
Haslam Jane
Lyons Alexander Ralph
Rowe Simon Michael
Taylor Richard Ian
Kassa Yosef
Mehta Bhavesh
LandOfFree
Image processing apparatus does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Image processing apparatus, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Image processing apparatus will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3144185