Device for transcribing images on a board using a camera based board scanner

• US 5,528,290 A
• Filed: 09/09/1994
• Issued: 06/18/1996
• Est. Priority Date: 09/09/1994
• Status: Expired due to Term

• Current Assignee: Xerox Corporation (Xerox Holdings Corp.)
• Original Assignee: Xerox Corporation (Xerox Holdings Corp.)
• Inventors: Saund, Eric
• Primary Examiner(s): GARBER, WENDY RAE
• Application Number:
  US08/303,918
• Time in Patent Office: 648 Days
• Field of Search: 348/218, 348/143, 348/159, 348/362, 348/373, 382/41, 382/44, 382/45, 382/284, 382/293, 382/295
• US Class Current: 348/218.1
• CPC Class Codes: G06V 10/10 Image acquisition document image scanning and transmission H04N1/00; control of digital cameras H04N23/60

Abstract

A device for transcribing, into electronic form, markings drawn on a whiteboard or blackboard. An electronic camera such as an ordinary video camera is mounted on a computer-controlled pan/tilt head in the ceiling or to the side of the board. Images are captured by directing the camera successively at small regions of the board, then processing these snapshots electronically, leading to a complete, undistorted, high-resolution image of the entire board.

Claims

1. An automatic system for capturing in computational form the image content of an image on a surface, comprising;

a) a surface upon which is provided an image;

b) at least one camera subsystem for capturing said image in a matrix of image tiles, each of said tiles comprising at least a portion of the image, said tiles each overlapping adjacent tiles, at least one of said tiles including perspective distortion of the portion of the image;

c) a processor for receiving and processing said image tiles to combine said tiles to determine a complete computational image corrected for perspective distortion representing said image, the processor determining an overlapping area in each tile overlapping adjacent tiles, finding a mark in each of said adjacent tiles representing a mark in the overlapping area, and combining said tiles based on said mark to produce a single computational image corrected for perspective distortion representing said image, the processor combining said tiles to produce a single computational image corrected for perspective distortion by;

1) for each two adjacent tiles, determining a region of overlap between a first tile and a second tile;

2) determining at least a mark in said region;

3) finding a projection of said at least one mark at a first surface coordinate location in said first tile image;

4) finding a projection of said mark at a second surface coordinate location in said second tile;

5) correcting perspective distortion between said first and second surface coordinate locations; and

6) writing corrected surface coordinates into a representation of said computational image stored in said memory; and

d) memory accessible by said processor for storing data including instructions for said processor and said computational image corrected for perspective distortion.

2. The system of claim 1, wherein said processor corrects perspective distortion of said surface coordinates of said mark by:

computing a first perspective transform of said first tile to obtain a first set of transformed surface coordinates according to said first transformed tile;

computing a second perspective transform of said second tile to obtain a second set of transformed surface coordinates, said second perspective transform weighted by a confidence of said second surface coordinate location according to said second tile;

determining a difference between said first set of transformed surface coordinates and said second set of transformed surface coordinates;

wherein said processor repeats said steps of computing a perspective transform of said second tile and determining a difference of said first set of and said second set of transformed surface coordinates until said difference is less than an acceptable threshold.

3. The system of claim 2, wherein the processor further determines a confidence of said second surface coordinate location being in a determined position to said second tile by

determining an image patch in an area surrounding and including at least a portion of said mark in said first tile;

correlating said image patch with said mark in said second tile;

determining a confidence area in said second tile in which said image patch correlates with said mark; and

determining a lengthwise confidence in a lengthwise orientation of said mark, and a widthwise confidence in a widthwise orientation of said mark;

said lengthwise confidence and said widthwise confidence being used to weight said second perspective transform.

4. The system of claim 1, wherein said processor processes said tiles by correcting for lightness variation across each tile, determining area in each tile overlapping surrounding tiles, combining said tiles to produce a single intermediate grey-level image, and thresholding said grey-level representation of said image to produce said binary image representing said image.

5. The system of claim 4, wherein said processor combines said tiles to produce a single intermediate grey-level image by:

for each two neighboring tiles, determining a region of overlap between a first tile and a second tile;

determining at least a mark in said overlapping region;

finding a projection of said mark at a first surface coordinate location in said first tile;

finding a projection of said mark at a second surface coordinate location in said second tile;

correcting perspective distortion between said first and second surface coordinate locations; and

writing corrected surface coordinates into said intermediate image representation stored in said memory.

6. The system of claim 1, wherein said at least one camera subsystem comprises a single camera rotatable with respect to said surface.

7. The system of claim 1, wherein said at least one camera subsystem comprises a plurality of cameras each in a stationary position with respect to said surface, each scanning an image tile of said surface.

8. The system of claim 1, wherein said at least one camera subsystem comprises a plurality of cameras each in a stationary rotatable position with respect to said surface, each scanning at least one image tile of said surface.

9. A method for providing a complete undistorted computational representation of an image on a surface, the method comprising the steps of:

a) capturing image tiles using at least one camera subsystem, each tile containing a portion of the image, said tiles each overlapping adjacent tiles;

b) converting said image tiles into electronically computational format;

c) determining said overlapping area in each tile;

d) correcting for distortions including perspective distortion in said overlapping area to produce a single computational image by the substeps of;

1) determining landmarks falling in said area of overlap;

2) performing local correlations of image patches in neighborhoods of landmarks to determine corresponding landmark locations in each image tile;

3) projecting said corresponding landmark locations into surface coordinates;

4) comparing pairs of landmarks from overlapping tiles;

5) choosing target coordinates at intermediate locations; and

6) computing new perspective transforms of each tile for said intermediate locations;

wherein said substeps are repeated until convergence upon landmark positions between overlapping tiles represented by cumulative errors in landmark projection being less that a threshold amount.

10. A method for providing a complete undistorted computational representation of an image on a surface, comprising:

a) capturing image tiles using at least one camera subsystem, each tile containing a portion of the surface image, said tiles each overlapping adjacent tiles;

b) converting said image tiles into electronically computational format;

c) correcting for lightness variation across each tile;

d) determining said overlapping area in each tile;

e) correcting for distortions including perspective distortions in said overlapping area to produce a single intermediate grey-level image by the substeps of;

1) determining landmarks falling in said region of overlap;

2) performing local correlations of image patches in neighborhoods of landmarks to determine corresponding landmark locations;

3) projecting said corresponding landmark locations into surface coordinates;

4) comparing pairs of landmarks from overlapping tiles;

5) choosing target coordinates at intermediate locations; and

6) computing new perspective transforms of each tile for said intermediate locations;

wherein said substeps are repeated until convergence upon landmark positions between overlapping tiles represented by cumulative errors in landmark projection being less that a threshold amount; and

f) thresholding said grey-level representation of said image to produce an computational image without perspective distortion representing said image.