WO2001061628A1

WO2001061628A1 - Apparatus and method of finding active microfiche image regions on a carrier

Info

Publication number: WO2001061628A1
Application number: PCT/US2000/010163
Authority: WO
Inventors: Marek A. Niczyporuk; Glenn S. Kimball
Original assignee: Niczyporuk Marek A; Kimball Glenn S
Priority date: 2000-02-15
Filing date: 2000-04-14
Publication date: 2001-08-23
Also published as: AU2000244618A1

Abstract

A low-resolution prescan camera (18) having a wide field of view captures a series of images indicative of a microfiche image carrier having a plurality of microfiche images. A stitching algorithm (300) stitches the series of images together to form a low-resolution image of the microfiche image carrier and the various ones of microfiche images disposed thereon. A find active image area algorithm (400) determines the location and size of each individual one of the microfiche images for subsequent conversion into corresponding digital images.

Claims

We claim: 1. A method (100) of identifying the active image regions on a light passing document, comprising: locating (300, 400, 408) individual ones of a plurality of uniformly distributed rows of active image areas disposed on the light passing document; and locating ( 300, 400, 412) individual ones of a plurality of non uniform image regions disposed within each individual one of the identified row locations.

2. The method (400) according to claim 1 , wherein said step of locating individual ones of a plurality of uniformly distributed row of active image areas includes: scanning (300, 306, 308, 312) the light passing document with a low resolution camera (16) to form a sequence of images indicative of the light passing document; and stitching together (309, 310) said sequence of images to form prescan image data indicative of the light passing document and individual ones of said plurality of uniformly distributed rows of active image areas.

3. The method (100) according to claim 2, wherein said step of locating individual ones of a plurality of uniformly distributed row of active image areas includes: fitting (400, 404, 406, 408) a uniform horizontal line array to the prescan image data to define a quality-of-fit.

53

4. The method (100) according to claim 3, wherein said step of fitting (400, 404, 406, 408) a uniform horizontal line array to the prescan image data including defining a quality-of-fit variable Q-Fit for fitting uniform horizontal row patterns, said quality-of-fit variable being defined by: Q-Fit = Q-FitLin +

Q-FitFuzzy.

5. The method (100) according to claim 4, wherein Q-FitLin is a linear quality-of-fit variable using a plurality of image intensity-based components.

6. The method (100) according to claim 5, wherein said plurality of image intensity-based components includes: a polarity signed positive or negative to maximize the differences between intensity along borderlines as opposed to mid-image intensity between borderlines; a negative absolute difference between an average intensity along a row border (IAveLin) between consecutive rows (n, n+1) to minimize the differences between intensity along consecutive horizontal row borderlines; and a positive variance given by the difference between a variance of horizontal pixel-to-pixel intensity within a mid-image region between row borders (IvarMid) and a variance along row borderlines (IvarLin), where positive variance expects high variance along active image areas and low variance along horizontal row borderlines.

7. The method (100) according to claim 6, wherein Q-FitFuzzy is a fuzzy quality-of-fit variable using another plurality of image intensity-based components.

8. The method (100) according to claim 4, wherein said another plurality of

54 image intensity-base components includes: a set of components of average intensity and variance along a wide-fuzzy horizontal row-border line (IAveLin, IVarLin), wherein maximum intensity is utilized in a small vertical neighborhood near an ideal horizontal line.

9. The method (100) according to claim 1, wherein said step of locating

(300, 400, 408) individual ones of a plurality of uniformly distributed row of active image areas includes: enhancing a pre-scanned image indicative of the light passing document using a predefined operator wherein said predefined operator is a BLENDED- MAX-N operator; fitting a maximized predefined variable to the enhanced pre-scanned image, wherein said predefined variable is a Q-FitPos variable; enhancing said pre-scanned image using another predefined operator, wherein said another predefined operator is a BLENDED-MIN-N operator; fitting another maximized predefined variable to the last mentioned enhanced scanned image, wherein said another maximized predefined variable is a Q-FitNeg variable; and selecting the lεirger between the Q-FitPos fitted image and the Q-FitNeg image to identify the uniform active image rows in said pre-scanned image.

10. The method (100) according to claim 1, wherein said step of locating (300, 400, 412) individual ones of a plurality of non uniform image regions disposed within each individual one of the identified row locations includes: defining a quality-of-fit variable Q-Reg for fitting rectangular regions to image data within a determined row.

55

11. The method (100) acording to claim 10, wherein said quality-of-fit variable Q-Reg is defined as a linear combination of image intensity-based components including: a signed polarity (IAveLin - IAveMid) defined as a difference between an average intensity along an edge line of a region (IAveLin) and an interior of said region (IAveMid), said sign of polarity indicating if said borders are expected brighter or darker than the active mid-image region to facilitate maximizing the difference between intensity along borderlines of said region and the mid-image intensity of said region; and a negative absolute (-abs (IaveLinl - IAveLin2)) wherein said negative absolute is the difference between the average intensity between two region edge lines, either in a horizontal or vertical direction to facilitate maximizing uniformity of image intensity along borderlines.

12. The method (100) according to claim 1 , wherein said step of locating (300, 400, 408) individual ones of a plurality of uniformly distributed row of active image areas includes: enhancing a pre-scanned image indicative of the light passing document using a predefined operator wherein said predefined operator is a BLENDED- MAX-N operator; fitting a maximized predefined variable to the enhanced pre-scanned image, wherein said predefined variable is a Q-FitPos variable; enhancing said pre-scanned image using another predefined operator, wherein said another predefined operator is a BLENDED-MIN-N operator; fitting another maximized predefined variable to the last mentioned enhanced scanned image, wherein said another maximized predefined variable is a Q-FitNeg variable; and selecting the larger between the Q-FitPos fitted image and the Q-FitNeg image to identify the uniform active image rows in said pre-scanned image.

56

13. The method (100) of processing discrete images, comprising: determining (300, 400) the area location of individual ones of a plurality of discrete image areas arranged in uniform rows and non uniform columns on an image bearing substrate; and scanning (500) only the determined area location of each individual one of said plurality of discrete images disposed on said image bearing substrate.

14. A method (100) of processing discrete images, comprising: fitting optimal row geometry with maximized Q-Ft statistic to establish the location of at least one row of discrete images, wherein said at least one row of discrete images has an established row height; and fitting simple active image regions in their respective locations in said at least one row of discrete images, wherein each simple active image region has an assumed height and a determine width with no space between any adjacent active image region in said at least one row of discrete images.

15 The method (100) of processing discrete images according to claim 14, hwherein said assumed height matches said established row height.

16. The method (100) of processing discrete images according to claim 14, wherein said step of fitting active image regions is an iterative process.

17. The method (100) of processing discrete images according to claim 14, further comprising converting each simple active image region to a corresponding 2-dimensional region.

18. The method (100) of processing discrete images according to claim 17, further comprising classifying each simple active image region.

19. The method (100) of processing discrete images according to claim 18, wherein said step of classifying includes: classifying false border regions; classifying very lεirge regions, where a very large region is defined when the width of said very large area is significantly larger than the average width of the remaining images within said at least one row; classifying very small regions, where very small regions are classified as false border regions;

57 classifying empty space near a first image or a last image in said at least one row; and normalizing and refitting optimal region sizes for each active region to maximize quality-of-fit for each image separately.

20. An apparatus (10) for identifying the active image regions on a light passing document, comprising: means (14, 100, 300, 400) for locating individual ones of a plurality of uniformly distributed rows of active image areas disposed on the light passing document; and means (14, 100, 300, 400) for locating individual ones of a plurality of non uniform image regions disposed within each individual one of the identified row locations.

58