Image analysis – Image segmentation
Reexamination Certificate
2000-08-14
2004-07-20
Boudreau, Leo (Department: 2621)
Image analysis
Image segmentation
C382S280000, C348S586000, C348S587000, C348S590000, C348S591000
Reexamination Certificate
active
06766054
ABSTRACT:
FIELD OF THE INVENTION
The present invention is directed to digital photography. It is more particularly directed to separating objects of interest from the background of a photograph or digitized image.
BACKGROUND OF THE INVENTION
The field of digital photography is growing rapidly because of the availability of inexpensive digital cameras and low cost personal computers to process the digitally acquired images. One of the areas where digital photography is seeing increased use in studio photography, where objects of interest are placed in a controlled environment. This environment includes specific studio lighting and backgrounds. Since an image of the object is acquired digitally rather than on film, this digital image is subject to several digital image processing techniques on a computer to improve its appearance. This is advantageous in that the processing can be performed almost instantaneously without the many steps required with film-based media.
A frequently used processing technique is that of object-background separation. This results in the separation of those image pixels belonging to the object of interest from the background. Since this separation takes place with a visual image of the object, rather than a physical separation of the object from the background, we refer to this as visual separation. Visual separation frees the image of the object from the background, so that the object can be superimposed on other backgrounds at a later time. For instance, it may be desirable to place the objects on white backgrounds for printed catalogs, whereas it may be desirable to place the objects on a black background for web-based publishing. Isolating an object from its background gives one flexibility to change the background as desired.
Methods for separating an object from its background generally define characteristics or properties of the object and background that are distinct from one another. For instance, color is one such characteristic or property. A commonly used technique to perform object-background separation is that of chroma keying. This technique groups areas within an image based on color similarity. More generally, assume the digital image is represented in one of the popular color spaces: RGB, YCrCb, LAB, LUV, HSB, CMY, or CMYK. Then the generalization of the chroma-keying idea is to segment the image into selected pixels that lie within a fixed distance from a distinguished point in this space, and into non-selected pixels that have greater distance from the distinguished point. Many popular image processing programs provide tools that operate in this manner. In one case, this tool not only allows for the selection of the distinguished point in the color space and the distance from that point, but optionally refines the segmentation by adding the constraint that all selected pixels must be contiguous. This requires that related pixels must either be neighbors of a selected pixel, or be neighbors of defined neighbors of that pixel, etc.
There are several disadvantages in the practical use of chroma-keying. Lighting variations and shadows can cause errors in the measurement of color. Furthermore, if the object is shiny, it can reflect the color of the background, causing mislabeling of object pixels.
It is advantageous to have a method that overcomes these and other limitations of chroma-keying. Manual segmentation, in which a human painstakingly outlines the object is a time consuming process.
Texture is a characteristic or property of an object or background. As used herein, image texture is defined as ‘the surface markings or two dimensional appearance of a surface’. It is well known that image texture (or texture characteristic) is an important image characteristic used by the human visual system as a cue in differentiating objects from each other. In fact, if the image is monochromatic, such as black and white, texture provides significant information in separating different objects from each other. Other cues such as continuity and perceptual grouping are also used by the human visual system, but these are typically higher level processes. There are few computational techniques that use such higher level cues in a robust, repeatable fashion. On the other hand, image texture has been widely studied from a computational standpoint and there are several image texture measures that work well in practice.
A texture discrimination method which separates characters from background in the headline areas of documents is limited to character recognition. A system for weed recognition and identification utilizes image segmentation. The segmentation is performed by means of a sensor to detect green vegetation and classify green image regions based on their shape and texture. Another method for segmenting an object from its background using texture information performs segmentation using entropic thresholding.
It would be quite advantageous to be able to perform object-background separation in a way that does not depend on color information alone.
SUMMARY OF THE INVENTION
It is therefore an object of the present invention to provide methods and apparatus for object-background separation using texture information and/or characteristics.
An aspect of the invention provides a method for visual separation. The method includes providing a digitized image of at least one object upon a textured background, and employing a texture characteristic in separating at least one object in the image from the textured background and/or background region.
Another aspect of the invention is provision of an apparatus for separating at least one object in the image from the textured background. In a particular embodiment, the apparatus includes a digital camera system; a lens optically coupled to the digital camera system for capturing a digitized image. The digital camera has an output coupled to a computer. The computer having code of a program for effecting the visual separation of at least one object upon a textured background included in the digitized image.
REFERENCES:
patent: 5140416 (1992-08-01), Tinkler
patent: 5345313 (1994-09-01), Blank
patent: 5448651 (1995-09-01), Sakou et al.
patent: 5481095 (1996-01-01), Mitsuda et al.
patent: 5502482 (1996-03-01), Graham
patent: 5561611 (1996-10-01), Avinash
patent: 5606821 (1997-03-01), Sadjadi et al.
patent: 5699452 (1997-12-01), Vaidyanathan
patent: 6034740 (2000-03-01), Mitsui et al.
patent: 6271890 (2001-08-01), Tamir et al.
patent: 6307975 (2001-10-01), Hosaka et al.
patent: 6556784 (2003-04-01), Onuki
Cutler, R. and Davis,L.; View-Based Detection and Analysis of Periodic Motion, Pattern Recognition, 1998. Proceedings. Fourteenth International Conference on, vol.: 1, Aug. 16-20, 1998, pp. 495-500.
Christensen James Edward
Cole Alan George
Rao Ravishankar
Akhavannik Hussein
Boudreau Leo
Herzberg L. P.
Petraske E. W.
LandOfFree
Segmentation of an object from a background in digital... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Segmentation of an object from a background in digital..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Segmentation of an object from a background in digital... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3219959