Image analysis – Image segmentation – Distinguishing text from other regions
Reexamination Certificate
2000-02-28
2003-09-23
Mehta, Bhavesh M. (Department: 2625)
Image analysis
Image segmentation
Distinguishing text from other regions
C382S180000, C358S002100, C358S462000
Reexamination Certificate
active
06625312
ABSTRACT:
BACKGROUND OF THE INVENTION
The present invention generally relates to a method and apparatus to improve the appearance of scanned images rendered on a recording medium. More particularly, the present invention is directed to automatically classifying the content of a scanned image based on segmentation tags statistics.
In the reproduction or display of images, and more particularly, the rendering of image data representing original document that has been electronically scanned, one is faced with the limited resolution capabilities of the rendering system. Tailoring an image processing system to offset the limitations of the rendering system is difficult due to the divergent processing needs required by different image types. Optimizing a system for one image type typically comes at the expense of degraded rendering of other image types. For example, optimizing the system for reproduction of continuous tones (contones) or photographic images often comes at the expense of degraded rendering of text and line art. Further complicating the reproduction of original documents is the reality that a document may be comprised of multiple image types (image classes).
To address this situation, digital reproduction devices often use automatic image segmentation techniques. Auto-segmentation is a well known operation that may use any of a number of classification functions (e.g., auto-correlation, frequency analysis, pattern or template matching, peak/valley detection, histograms, etc.) to analyze video image data and classify image pixels as one of several possible image classes based on the pixel characteristics. Auto-segmentation processes typically generate a pixel classification signal, known as a segmentation tag, that identifies the pixel as a particular image class. A one-pass digital reprographic system (scanning and printing done in a single pass of the image) gets just one chance to analyze and classify each pixel of an image based on several pixels from a few scanlines of neighboring data. Due to the limited context for classification one-pass auto-segmentation can include erroneous and/or abrupt switching between image classes which, in turn, may result in the formation of visible artifacts in the resulting output image.
One approach to improving image classification in digital reproduction systems is the use of an image classification system which utilizes a fuzzy membership into each category or class. In a fuzzy classification system the classes are not mutually exclusive, thereby eliminating problems with class switching and also allowing those areas to have processing different than that of any of the other pre-defined classes; i.e., the output can choose between a continuum of possible image processing techniques. While fuzzy classification hides or eliminates most misclassification, some users find that the overall image quality is compromised.
Another approach to improve the reproduction of scanned images is to employ a two-pass auto-windowing process. In a two-pass process, the scanned image data is analyzed twice. The first-pass is used to automatically locate “windows” contained within a scanned document image by identifying image runs within a scanline and combining image runs from adjacent scanlines into windows. In the second-pass, the pixels within each window are classified and labeled as a particular image type. A two-pass auto-windowing process provides good image quality and avoids most of the switching artifacts associated with single pass system. However, two-pass systems perform a large number of intensive calculations and require buffering of tags, thus making them much more costly and complex than single pass systems.
Yet another approach to improving the reproduction of images is to provide multiple processing modes, each optimized for a different type of original. Most digital copiers provide for optimized processing based on the content of an original document by including user selection of one of multiple copy modes such as, for example, modes for text documents, photographs, mixed content, etc. in addition to an automatic mode. This selection allows a user to choose alternative modes to avoid defects that the user may perceive with the automatic mode. However, such a system requires the user to be familiar with the characteristics of each image type to properly identify the proper mode. Without such familiarity, users often must make and compare copies from different modes to obtain the best image quality for a given document. Furthermore, the automatic mode in such systems typically makes use of one of the segmentation options identified above. Thus, such devices suffer from the limitations and drawbacks identified above.
Therefore, it is desirable to have a system that minimizes segmentation artifacts and provides optimum image quality.
The following references may be found relevant to the present disclosure:
U.S. Pat. No. 5,327,262 to Williams which discloses in conjunction with an image segmentation arrangement in which an image is processed with an image type detection arrangement, a morphological filtering operation which initially provides a noise removal filter operating the image detection signal to remove noise within an area of the image detection signal and subsequently provides a hole filling filter which bridges small gaps in the image type detection results.
U.S. Pat. No. 5,765,029 to Schweid et al. which discloses a method and system that electronically fuzzy classify a pixel belonging to a set of digital image data with respect to a membership of the pixel in a plurality of image classes. This process determines a fuzzy classification of the pixel and generates an effect tag for the pixel based on the fuzzy classification determination.
U.S. Pat. No. 5,778,156 to Schweid et al. discloses system and method that electronically image process a pixel belonging to a set of digital image data with respect to a membership of the pixel in a plurality of image classes. This process uses classification to determine a membership value for the pixel for each image classes and generates an effect tag for the pixel based on the fuzzy classification determination. The pixel is image processed based on the membership vector of the pixel.
U.S. Pat. No. 5,850,474 to Fan et al. discloses a method and apparatus for segmenting image data into windows and for classifying the windows as typical image types which include making two passes through the image data. The method includes a step of making a first pass through the image data to identify windows and to record the beginning points and image types of each of the windows, and a step of making a second pass through the image data to label each of the pixels as a particular image type.
U.S. Pat. No. 5,852,678 to Shiau et al. discloses a method and apparatus for improving digital reproduction of a compound document image containing half-tone tint regions and text and/or graphics embedded within the half-tone tint regions. The method entails determining a local average pixel value for each pixel in the image, then discriminating and classifying based on the local average pixel values, text/graphics pixels from half-tone tint pixels.
SUMMARY OF THE INVENTION
One aspect of the present invention is a method of classifying image data based on tag statistics and rendering the image data accordingly to provide a favorable image quality. The method includes receiving segmentation tags for a block of scanned image data; determining an image type for the block of scanned image data as a function of the segmentation tags; and generating rendering tags pixels within the block of scanned image data as function of the image class for the pixel and the image type.
Another aspect of the present invention is a system for classifying image data based on tag statistics. The system includes a segmentation processor generating segmentation tags for pixels within a block of scanned image data; a tag analyzer for collecting tag statistics for the segmentation tags; a scanning processor for determining an image type for the
Li Xing
Nagarajan Ramesh
Tse Francis K.
Eipert William
Mehta Bhavesh M.
Sukhaphadhana Christopher
Xerox Corporation
LandOfFree
Document classification using segmentation tag statistics does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Document classification using segmentation tag statistics, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Document classification using segmentation tag statistics will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3112549