Computer graphics processing and selective visual display system – Computer graphics processing – Attributes
Reexamination Certificate
2001-07-12
2004-10-05
Razavi, Michael (Department: 2672)
Computer graphics processing and selective visual display system
Computer graphics processing
Attributes
Reexamination Certificate
active
06801210
ABSTRACT:
BACKGROUND OF THE INVENTION
1. Field of the Invention
The present invention relates to the field of image representation. More particularly, the present invention relates to a method of representing images using mathematical models of geometric and brightness characteristics of the image, known as Content Oriented Representation (“CORE”).
2. Background Information
Analysis and processing of digital images play an important role in nearly all disciplines of modern industry and economic activity. From medical imaging to industrial quality control and diagnostics, to entertainment and advertising, efficient image analysis, representation and processing is the primary component of overall imaging system performance.
Presently, there are forms of content-oriented representation of images known in the field of image representation. Partial implementations exist, known generally as “vector formats” and “vectorizations,” which are representations of visual images by geometric entities, such as vectors, curves, and the like. Usually, vectorized images are significantly more compact and easier to process than identical images represented by conventional techniques relying on use of pixels, for example.
Currently available products incorporate limited vector formats, including for example, “Photoshop,” developed by Adobe Systems Incorporated, “Flash”and “Shockwave,” developed by Macromedia, Inc., and W3C Scalable Vector Graphics (SVG). However, vectorization methods employed by these products provide cartoon-like images and animations. In other words, they fail to adequately represent high resolution, photo-realistic images of the real world. This is because only simple, cartoon-like images allow for representation by edge partitions, which are not necessarily present in typical photo-realistic images. In contrast, high resolution, real world pictures present an enormous variety of forms and highly complex visual patterns, which conventional vectorization methods fail to capture. In fact, high resolution, real world images present such an enormous variety of forms and complex visual patterns, that visually accurate vectorization is practically impossible under the existing methods.
Existing vectorization techniques are confined by certain limitations, which must be overcome to adequately provide content-oriented representation of high resolution photo realistic images. The basic requirements of effective image representation include the following: (i) the resulting image has no visible distortions; (ii) the number of parameters in the resulting image is an order of magnitude less than the number of pixels in the original image; (iii) the parameters have simple visual interpretations; (iv) all reasonable image transformations are expressible in terms of the representation parameters, so that all the image processing operations are possible; and (v) with respect to video sequences and video compression, subsequent frames of the resulting image behave coherently, such that the models remain basically the same, while only respective geometric parameters change continuously.
Although the existing methods of image representation, processing and compression, such as DCT transform and the JPEG compression standard, as well as various wavelets transforms and compression schemes, may satisfy the first requirement above, they fail with respect to the remaining four requirements. Current methods of image representation are based on linear transformations of the image to a certain basis, which contains initially the same number of elements as the number of pixels in the original image. Subsequent quantization and filtering reduces the number of parameters, but in an unpredictable fashion. Also, visual interpretation of the reduced number of parameters may be difficult.
Moreover, because video sequences represent exactly the motion of certain objects and patterns (i.e., geometric transformations of the initial scene), the DCT or the wavelets representations of video sequences behave in an incoherent and unpredictable manner. Therefore, existing video compression techniques, such as MPEG, use JPEG compression for the first frame and perform motion compensation on a pixel level, as opposed to a compressed data level. This results in a tremendous reduction in efficiency.
A method for image representation and processing is described by James H. Elder and Rick M. Goldberg in “Image Editing in the Contour Domain,” IEEE (1998), based on edge capturing, together with the “blur scale” and the brightness values. Although providing additional efficiency to image representation and processing in the “geometric image domain,” the disclosed method does not solve the main problems of the existing methods. In particular, the method relies on only edges, while ignoring more complicated characteristic lines. Likewise, the method ignores possible geometric proximities and crossings between edges. Reconstruction of brightness values between the edges relies on solving Laplace transform equations, which appears to be an ad hoc operation that does not take into account actual image brightness. Furthermore, the method does not include any tools for representing background and texture visual patterns. The geometric accuracy of the suggested edge detection method (i.e., marking the nearest pixel and edge direction) is not sufficient for a faithful image reconstruction. Lastly, the ad hoc Gaussian blur model does capture the actual cross-sections of the edges.
Advances in vectorization of high resolution images, however, have continued to evolve. For example, U.S. Pat. No. 5,410,643 to YOMDIN et al., the disclosure of which is expressly incorporated herein by reference in its entirety, describes a method for image data representation by mathematical models of a certain type. However, the visual quality and compression ratio is low and image processing on the compressed data is impractical. In U.S. Pat. No. 5,510,838 to YOMDIN et al., the disclosure of which is expressly incorporated herein by reference in its entirety, the images are represented by four types of models: edges, ridges, hills and background. Edges and ridges are represented by mathematical models that include polygonal lines representing the center lines of the edges or ridges and the corresponding brightness (or color) profiles. The brightness profiles are kept at the vertices of the polygonal lines and interpolated along segments of these lines. Hills, which correspond to small patches on the image, are represented by paraboloid-like mathematical models. Background is represented by low degree polynomials placed on a predetermined artificial grid.
More particularly, U.S. Pat. No. 5,510,838 discloses a method for detection of edges, ridges, hills and background, based on approximation of the image by second and third order polynomials on overlapping 4×4 and 5×5 pixels cells and further analysis of these polynomials. This method provides better quality and compression than the method disclosed in U.S. Pat. No. 5,410,643, for example. However, the image quality and resolution are not sufficient for most practical applications, the compression ratio is inferior to that of other conventional methods and processing the compressed data is cumbersome and complicated.
Other practical disadvantages of the invention disclosed in the U.S. Pat. No. 5,510,838 include the following: (i) the image is subdivided into cells of a size of 6 to 48 pixels, which cells are represented independently, thereby reducing image quality and the compression ratio; (ii) there is no accounting for visual adjacencies between the models; (iii) approximation of edges and ridges by polygonal lines causes visual degradation (e.g., a “staircase effect”); (iv) the resolution of the detection method is insufficient due to second degree polynomial approximations on 4×4 pixel windows; (v) representation of the background is unstable and inefficient, resulting in visual degradation of the image, low compression and cumbersome processing; and (vi) the only tool for representing backgro
Elichai Yoram
Yomdin Yosef
Fenster & Company
Good-Johnson Motilewa
Razavi Michael
Vimatix (BVI) Ltd.
LandOfFree
Method and apparatus for image representation by geometric... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Method and apparatus for image representation by geometric..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Method and apparatus for image representation by geometric... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3290347