System and method for identifying and classifying color regions from a digital image

• US 8,244,031 B2
• Filed: 04/14/2008
• Issued: 08/14/2012
• Est. Priority Date: 04/13/2007
• Status: Active Grant

• Current Assignee: Tungsten Automation Corp.
• Original Assignee: Kofax Incorporated
• Inventors: Cho, Seungyeon, Bither, William
• Primary Examiner(s): Seth, Manav
• Application Number:
  US12/102,419
• Publication Number:
  US 20080253647A1
• Time in Patent Office: 1,583 Days
• Field of Search: 382100-101, 382104-105, 382112-113, 382162-231
• US Class Current: 382/162
• CPC Class Codes: G06V 30/10 Character recognition
• CPC Class Codes: G06V 30/18105 related to colour
• CPC Class Codes: G06V 30/413 Classification of content, e.g. text, photographs or tables

Abstract

A system and method of identifying and classifying regions of a digital image. The method includes an initial step of inputting an image as a color digital image. Subsequently, information that identifies color regions of the color digital image is obtained. Finally, color and non-color regions of the color digital image are classified based upon the identifying information.

Claims

1. A method of identifying and classifying regions of a digital image, said method comprising the steps of:

inputting an image as a color digital image;

obtaining information that identifies color and photo regions of said color digital image;

classifying color, photo, non-color and non-photo regions of said color digital image based upon said identifying information; and

individually compressing said color, non-color, photo and non-photo regions based on a quality characteristic of each of said regions;

wherein said photo regions are classified by the following steps;

transforming said color digital image to gray scale;

applying a first median filtering method to said gray scale image creating a first filtered image;

applying a second median filtering method to said gray scale image creating a second filtered image;

creating a difference image from between said first and second filtered images;

converting said difference image to a binary image;

obtaining a set of bounding boxes of black pixels in said binary image through a connected components method; and

applying a size analysis to said connected components wherein connected components having a size below a predetermined threshold are marked as non-photo regions and remaining bounding boxes of said connected components are photo regions.

2. A method of identifying and classifying regions of a digital image, said method comprising the steps of:

inputting an image as a color digital image;

obtaining information that identifies color and photo regions of said color digital image;

classifying color, photo, non-color and non-photo regions of said color digital image based upon said identifying information;

individually compressing said color, non-color, photo and non-photo regions based on a quality characteristic of each of said regions;

assessing a bit-depth of said color digital image to determine whether said image is entirely black and white or entirely grayscale; and

terminating said classification of said regions if said image has a bit-depth indicating that it is black and white or grayscale.

3. A method of identifying and classifying regions of a digital image, said method comprising the steps of:

inputting an image as a color digital image;

obtaining information that identifies color and photo regions of said color digital image;

classifying color, photo, non-color and non-photo regions of said color digital image based upon said identifying information;

individually compressing said color, non-color, photo and non-photo regions based on a quality characteristic of each of said regions; and

resampling and resizing said image to expedite the identification and classification of said regions of said digital image.

12. The method of claim 3 including the step of:

restoring said image to its original size prior to said resampling.

4. A method of identifying and classifying regions of a digital image, said method comprising the steps of:

inputting an image as a color digital image;

obtaining information that identifies color and photo regions of said color digital image;

classifying color, photo, non-color and non-photo regions of said color digital image based upon said identifying information;

individually compressing said color, non-color, photo and non-photo regions based on a quality characteristic of each of said regions;

identifying black border regions in said image; and

removing any identified black border regions.

5. The method of claim 4 wherein said identification of black border elements includes the steps of:

determining whether pixels that comprise said color digital image are black;

scanning said black pixels along a dimension of said image to determine a boundary of said black border;

establishing said boundary of said black border where said non-black pixels exceed one-percent of a total of said pixels of said scanned dimension.

6. A method of identifying and classifying regions of a digital image, said method comprising the steps of:

inputting an image as a color digital image;

obtaining information that identifies color and photo regions of said color digital image;

classifying color, photo, non-color and non-photo regions of said color digital image based upon said identifying information;

individually compressing said color, non-color, photo and non-photo regions based on a quality characteristic of each of said regions;

identifying gray elements in said image; and

removing any identified gray elements.

7. The method of claim 6 wherein identifying gray elements includes:

transforming said color digital image to gray scale to create a gray scale image;

computing a Malahanobis distance for each color pixel in said color digital image to a corresponding pixel of said transformed gray scale image; and

marking a pixel in said color digital image as gray if it exceeds a predetermined distance from its counterpart in said gray scale image.

8. The method of claim 6 wherein said identification of color elements includes the following:

creating a binary image from said color digital image;

marking any pixels previously identified as gray or photo elements as white and any remaining pixels as black;

applying a connected components method to obtain bounding boxes of connected components of said black pixels;

merging any overlapping bounding boxes into a single bounding box;

applying a size analysis method to said bounding boxes in which connected components having a bounding box of a smaller size than a predetermined value are marked as non-color and a remaining set of bounding boxes of said connected components are marked as a color region.

9. A method of identifying and classifying regions of a digital image, said method comprising the steps of:

inputting an image as a color digital image;

obtaining information that identifies color and photo regions of said color digital image;

classifying color, photo, non-color and non-photo regions of said color digital image based upon said identifying information;

individually compressing said color, non-color, photo and non-photo regions based on a quality characteristic of each of said regions;

identifying background elements; and

removing said background elements.

10. The method of claim 9 wherein identifying background elements includes:

creating a memory space having a width and height corresponding to that of said color digital image;

applying an HSL color transformation to said color digital image;

copying an H component from said HSL transformation for each pixel of said color digital image into said memory space;

applying an unsupervised clustering method to said H component data to identify background colors if present;

applying an image slicing method to said H component data to obtain a binary image in which background pixels are black;

obtaining a set of bounding boxes of using a connected components methodology for said black pixels; and

applying a size analysis to determine which bounding boxes are to be merged and considered background.

11. A method of identifying and classifying regions of a digital image, said method comprising the steps of:

inputting an image as a color digital image;

obtaining information that identifies color and photo regions of said color digital image;

classifying color, photo, non-color and non-photo regions of said color digital image based upon said identifying information;

individually compressing said color, non-color, photo and non-photo regions based on a quality characteristic of each of said regions; and

merging said photo and said color regions.

13. A digital imaging system comprising:

a color image capturing device configured for obtaining a digital image; and

a processor electrically connected to at least one of the image capturing device configured for obtaining information that identifies color and photo regions of said color digital image and classifying color, photo, non-photo and non-color regions of said color digital image based upon said identifying information, for individually compressing said color, non-color, photo and non-photo regions based on a quality characteristic of each of said regions, and for assessing a bit-depth of said color digital image to determine whether said image is entirely black and white or entirely grayscale and terminating said classification of said regions if said image has a bit-depth indicating that it is black and white or grayscale.