Method for automated image indexing and retrieval

1. A method of indexing images contained in scanned documents, wherein said scanned documents are stored in a repository, comprising:

for each document to be stored in the repository, dividing the document into a plurality of sections;

scanning the plurality of sections;

segmenting each scanned segment according to a predetermined coding model into image segment and non-image segments;

associating each of the image segments with the document;

generating an index correlating the image segments with the document;

defining the correlated image segments as index images; and

storing additional information about the index image.

2. The method of claim 1 , further comprising displaying the index of image segments in a user interface; and responsive to selection of an image segment from the index, displaying the document information associated with the image segment in the user interface.

3. The method of claim 1 , wherein the predetermined coding method comprises the multiple binary mask with multiple background window (MBM&MBW) model.

4. The method of claim 1 , wherein a section comprises one of a background windows and a binary mask layer.

5. The method of claim 4 , wherein prior to segmenting, excluding any uniform background windows.

6. The method of claim 5 , comprising defining an index image as an inter-correlated image segment in the background layer.

7. The method of claim 5 , comprising defining an index image as one of an inter-correlated image segment in the background layer and the background layer.

8. The method of claim 1 , wherein additional information pertaining to the index images includes type of image, size of image, color of image, and document page where the image occurs.

9. The method of claim 8 , wherein type of image is one of logo, photograph, graphic, bar graph, data plot, etc.

10. The method of claim 8 , wherein color of image includes hue and saturation.

11. The method of claim 1 , further comprising, for a multi-page document, for each image segment occurring on a page, associating the page number with the image segment; and adding an image of each page associated with the multi-page document to the index.

12. A method of indexing images contained in scanned documents, wherein said scanned documents are stored in a repository, comprising:

for each document to be stored in the repository, scanning the document to obtain a grayscale image;

determining at least one region of interest in the grayscale image;

determining predetermined statistical characteristics of the at least one region of interest;

based on the statistical characteristics of the at least one region of interest, classifying the at least one region of interest as either text or non-text;

associating each of the non-text regions of interest with the document; and

generating an index correlating the image segments with the document.

13. The method of claim 12 , wherein the statistical characteristics comprise aspect ratio, average hue and average saturation.

14. The method of claim 12 , wherein determining predetermined statistical characteristics of the at least one region of interest comprises determining a horizontal projection histogram.

15. The method of claim 12 , further comprising displaying the index of non-text region of interest in a user interface; and responsive to selection of a non-text region of interest from the index, displaying the document information associated with the non-text region of interest in the user interface.

16. The method of claim 15 , further comprising:

comparing data describing the region of interest's aspect ratio, average hue and average saturation to data describing a set of reference images aspect ratio, average hue and average saturation; and

if the data describing the region of interest's aspect ratio, average hue and average saturation is sufficiently similar to the data describing a reference image, displaying the document information associated with reference image in the user interface.

17. The method of claim 12 , wherein determining at least one region of interest in the grayscale image comprises:

determining a gradient matrix of the grayscale image on a pixel-by-pixel basis;

constructing a binary image matrix comprising those pixels in the gradient matrix having a value greater than a threshold value;

determining a cumulative gradient matrix for pixels within a predetermined distance;

constructing a binary image matrix of the gradient matrix comprising setting all pixels to one whose corresponding cumulative gradient value is non-zero; and

performing a connected component relationship on the cumulative binary image matrix to generate at least one region of interest.