Image analysis – Image compression or coding – Polygonal approximation
Reexamination Certificate
2000-11-21
2004-07-06
Mariam, Daniel (Department: 2623)
Image analysis
Image compression or coding
Polygonal approximation
C382S233000, C382S298000
Reexamination Certificate
active
06760483
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 identifying visual attributes of an image and transforming them into mathematical models.
2. Background Information
Analysis and processing of digital images plays an important role in nearly all disciplines of modem economics. From medical imaging to industrial quality control and diagnostics to entertainment and advertising, efficient image analysis, representation and processing is the primary component of the overall imaging system performance.
Some conventional methods of image analysis rely on extracting specific lines from the original images (referred to as “edges”), which separate distinct image areas of different brightness (or color). These specific lines may also be referred to as “ridges,” which are narrow strips, visually separated from the background by their brightness or color. Use of edge and ridge technology is described, for example, in relevant sections of A. Blake and M. Isard, “Active Contours” (1998); J. Foley, et al., “Computer Graphics: Principles and Practice” (1997); and R. Gonzales and P. Wintz, “Digital Image Processing” (1987).
Generally speaking, the importance of such “line-shaped” image patterns, both for human visual perception and efficient automated image analysis, processing and representation, is well recognized in the field. However, the existing methods provide only partial information on the geometric and brightness structures relating to the line-shaped patterns. This partial information is insufficient for adequate image analysis of patterns. As applied to accurate image reproduction and representation, the conventional methods based on edge/ridge partitions are inadequate, especially with respect to high resolution images of the real world.
There are two main reasons for the general failure of conventional methods. First, the edges and ridges are detected only as geometric lines on the image. They are formed by original pixels (or by pixels in a finer grid), or they are approximated by splines. Such limited geometric representation does not simultaneously capture brightness or color patterns along the detected lines. Without the brightness or color pattern information, visually authentic capturing and reconstruction of line-shaped “image forms” is not possible. Therefore, use of the conventional methodology in applications requiring realistic human visual perception of the image or high resolution analysis of medical, industrial and other commercially significant images is extremely limited.
The inadequate treatment of brightness patterns along line-shaped image forms results from difficulties associated with capturing such information. For example, whether or not a consistent brightness pattern exists along a given line is unpredictable. Typically, from a global perspective, both brightness and geometric shapes vary along any one line to such a large extent that direct comparison between any two distinct points on the line is virtually impossible. Furthermore, even for a line having somewhat uniform brightness and geometric characteristics, empirical determination of the corresponding brightness pattern is not possible by the conventional methods, because it requires explicit comparison of brightness data at different points along the line.
Second, edges and ridges as captured by conventional methods form only part of the visually significant line-shaped image patterns. Human visual experience demonstrates that much more complicated patterns than simply edges and ridges are perceived in line-shaped image forms. Conventional methods cannot recognize these more complicated forms.
In U.S. Pat. No. 5,510,838, to YOMDIN, et al., a method is disclosed for image representation by four types of models: edges, ridges, hills and background. Distinctive types of brightness profiles associated with the edge and ridge models are disclosed therein. However, the method of U.S. Pat. No. 5,510,838 presents a number of practical disadvantages. For example, the edge and ridge detection (and, in particular, detection of corresponding brightness profiles) is treated as an integral part of overall image representation. Consequently, isolated application of edge and ridge detection is difficult. Also, the method disclosed in U.S. Pat. No. 5,510,838 requires that the original image be subdivided into an artificial grid of cells, each having a size of 6 to 48 pixels. The cells in these grids must be processed independently. Such subdivision contradicts the global nature of edges and ridges, reduces the visual quality of the resulting model-based image and renders many image analysis and processing applications virtually impossible.
U.S. Pat. No. 5,510,838 has additional practical problems, common to conventional image representation methods. First, the shape of the brightness profiles is determined a priori, and permits only limited adjustment to the actual image pattern. Second, the method cannot be used for empirical determination of actual, more complicated brightness profiles present on the image. Also, the resolution of the detection method is generally limited due to the use of second degree polynomial approximations on 4×4 pixel windows.
Some of these problems are addressed in U.S. Pat. No. 5,960,118, to BRISKIN, et al. For example, an extended variety of brightness profiles is disclosed in this document. Also, the resolution of the detection method is improved by incorporating 3×3 pixel windows in the original polynomial approximation. The patent also describes a limited method for empirical profile detection, dependant on the assumption that the profile is constant along the subject line (which is not the case for most brightness profiles). Furthermore, U.S. Pat. No. 5,960,118 still does not allow for application of edge and ridge detection separately from the full image representation process. U.S. Pat. No. 5,960,118 also does not eliminate the necessity of subdividing images into cells.
SUMMARY OF THE INVENTION
In view of the above, the present invention through one or more of its various aspects and/or embodiments is presented to accomplish one or more objectives and advantages, such as those noted.
The present invention overcomes the shortcomings of the existing techniques of image analysis and processing based on edge and ridge detection. It provides a method for identification of general characteristic lines on an image, as well as the signatures and fitting transformations corresponding to each characteristic line. The signatures simultaneously capture the brightness (or color) pattern along detected lines and the fitting transformations consolidate similar brightness and geometric patterns, while maintaining visual integrity of the overall image.
The present invention describes a method for empirical determination of an actual brightness cross-section along any given line. The present invention also provides an efficient representation of characteristic lines by mathematical models. The method enables a number of image analysis implementations, including visual quality inspections, pattern detections and sensor calibrations, based on the analysis of characteristic lines and their signatures. Furthermore, all types of characteristic lines are accurately captured.
All aspects of the invention include steps that are implementable by a computing device, without input from a user (i.e., automatic), or by a computing device in combination with an interface device, with input from the user (i.e., interactive). The invention also includes any combination of automatic and interactive implementations.
An aspect of the invention provides for representing an image by identifying one or more characteristic lines of the image, each characteristic line being a line along which a visual pattern repeats itself. A central line, which corresponds to and approximates the shape of the characteristic line, is determined. A cross-section, consisting of one or more ge
Elichai Yoram
Yomdin Yosef
Fenster & Company
Mariam Daniel
Vimatix (BVI) Ltd.
LandOfFree
Method and apparatus for image analysis and processing by... 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 analysis and processing by..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Method and apparatus for image analysis and processing by... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3190854