Image analysis – Image segmentation – Distinguishing text from other regions
Reexamination Certificate
2000-05-26
2003-10-21
Mariam, Daniel G. (Department: 2621)
Image analysis
Image segmentation
Distinguishing text from other regions
C382S164000, C382S173000, C382S181000, C382S194000, C382S224000, C382S237000, C358S002100, C358S448000, C358S462000
Reexamination Certificate
active
06636630
ABSTRACT:
FIELD OF THE INVENTION
The present invention relates to an image-processing apparatus for use in digital copying machines, facsimile machines, etc. which carries out discrimination on character, photograph and mesh areas by extracting the amount of feature of each pixel from an image signal that has obtained by scanning a document.
BACKGROUND OF THE INVENTION
Conventionally, in digital copying machines and facsimile machines, in order to improve the picture quality of recorded images, image processing is carried out on image signals that have been obtained by reading a document having a character area, photograph area or mesh area or a document having a mixed portion of these areas, using a CCD (Charge Coupled Device) image sensor, or other devices.
The above-mentioned image processing includes a process for discriminating whether a read image belongs to a character, photograph or mesh image, and for carrying out an appropriate process for each pixel based upon the results of the discrimination. With respect to this type of image discrimination method for improving the image quality, examples thereof include a method in which an image is divided into blocks each having a plurality of pixels and each block is subjected to a pattern matching process, and a method in which the image discrimination is carried out on each block by using feature parameters representing characteristics of character images and mesh images. However, in the image discrimination method using the pattern matching process, many patterns have to be prepared, raising problems of an increase in memory capacity and degradation in general applicability. For this reason, in recent years, the image discrimination method using feature parameters has come to be widely used.
With respect to the image discrimination method using feature parameters, U.S. Pat. No. 4,722,008 (Date of Patent: Jan. 26, 1988) (corresponding to “Japanese Laid-Open Patent Application No. 194968/1986 (Tokukaishou 61-194968, published on Aug. 29, 1986)”) discloses a mesh-photograph area discrimination method. In this method, a document is divided into blocks, each having a plurality of pixels, and changes in signal levels in two continuous pixels located in a predetermined space within a block are measured in a separated manner between the case in which the two pixels continue in the main scanning direction and the case in which the two pixels continue in the sub scanning direction, and the sum of the measured values in each block is compared with each predetermined value, and the image is discriminated based upon the results of the comparison.
Moreover, with respect to another image discrimination method using feature parameters, Japanese Laid-Open Patent Application No. 147860/1987 (Tokukaishou 62-147860, published on Jul. 1, 1987) discloses a half-tone facsimile signal processing system. In this system, first, a difference between the maximum signal level and the minimum signal level within a block is found, and the value of the difference is compared with a preset value. In the case when the difference between the maximum signal level and the minimum signal level is smaller than the above-mentioned preset value, a judgment signal indicating a portion having a moderate signal level change that is supposed to be a picture portion is outputted. In contrast, in the case when the difference between the maximum signal level and the minimum signal level is greater than the above-mentioned set value, a judgment signal indicating a portion having an abrupt signal level change that is supposed to be an outline of a character and a photograph portion or a mesh-photograph portion. Moreover, in accordance with a predetermined accessing order within a block, the number of changes between the signal levels of respective two pixels that are spatially continuous is compared with a predetermined value, and in the case when the number of changes between the signal levels is greater than the predetermined value, a judgment signal indicating that the block in question belongs to a mesh area is outputted. In contrast, when the number of changes between the signal levels is smaller than the predetermined value, a judgment signal indicating that the block in question is not a mesh area is outputted. Thus, in response to these judgment signals, signal processing is carried out on pixels within each block.
Moreover, with respect to a conventional technique using maximum and minimum points among methods using feature parameters, Japanese Laid-Open Patent Application No. 14701/1993 (Tokukaihei 5-14701, published on Jan. 22, 1993) discloses an area discrimination device. In this area discrimination device, a threshold value is set based upon the difference between the maximum and minimum densities of pixels within a block consisting of a plurality of pixels. Then, the pixels within the block are successively scanned, and a change in the output amplitude, generated by density differences of the pixels, exceeds the above-mentioned threshold value, an output signal is generated. When the number of generations of the output signals exceeds a predetermined number of times, a judgment is made that the block in question belongs to a half-tone area.
Moreover, with respect to a conventional technique using another method in which the maximum and minimum points are adopted, Japanese Laid-Open Patent Application No. 152944/1994 (Tokukaihei 6-152944, published on May 31, 1994) discloses a mesh extracting device. In this mesh extracting device, digital multi-value data is inputted thereto, and a maximum point and/or a minimum point in density in the main scanning direction are/is detected and set as a max/min pixel(s), and in the case when the intervals of the max/min pixels are in a predetermined range and when the number of max/min pixels that exist between the max/min pixels and are projected in the sub scanning direction is not less than a predetermined number, these max/min pixels are detected as mesh pixels.
Furthermore, with respect to a conventional technique using another method in which the maximum and minimum points are adopted, Japanese Laid-Open Patent Application No. 178097/1994 (Tokukaihei 6-178097, published on Jun. 24, 1994) discloses a mesh extracting device. In this mesh extracting device, digital multi-value data is inputted thereto, and a maximum point and/or a minimum point in density is/are detected in the main scanning direction as a horizontal max/min points so that the number of pixels between the horizontal max/min points is counted. Moreover, based upon the above-mentioned multi-value data, a maximum point and/or a minimum point in density is/are detected in the sub scanning direction as a vertical max/min points so that the number of pixels between the vertical max/min points is counted. With respect to horizontal max/min points in which the number of pixels located between the horizontal max/min points is in a predetermined range, when the number of pixels located between the vertical max/min points is in a predetermined range, the corresponding pixels are detected as mesh pixels.
However, in the above-mentioned conventional image discrimination methods using feature parameters, an erroneous judgment as to image discrimination tends to occur. Therefore, in order to achieve high-quality images, further improvements in the discrimination precision have been demanded.
It is considered that the above-mentioned erroneous judgment in image discrimination is caused by inappropriate feature parameters, that is, the use of feature parameters that fail to properly represent characteristics of the respective areas. Moreover, another major cause of the erroneous judgment is that a classifying method, used for image discrimination with respect to the amount of feature obtained by a feature parameter, is not appropriate, and that the selection of threshold values for the classification is not appropriate. In particular, in the case when the maximum point and the minimum point are used as the feature parameters, inappropriate extracting methods for maximum and mini
Adachi Yasushi
Kanata Toshihiro
Conlin David G.
Edwards & Angell LLP
Hartnell, III George W.
Mariam Daniel G.
Sharp Kabushiki Kaisha
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-3140659