Devices and methods for restoring low-resolution text images

• US 8,437,551 B2
• Filed: 10/06/2008
• Issued: 05/07/2013
• Est. Priority Date: 10/05/2007
• Status: Expired due to Fees

• Current Assignee: Tufts University
• Original Assignee: Tufts University
• Inventors: Noonan, Joseph P., Basu, Prabahan
• Primary Examiner(s): Do, Anh Hong
• Application Number:
  US12/681,454
• Publication Number:
  US 20100208996A1
• Time in Patent Office: 1,674 Days
• Field of Search: 382/173, 382/176, 382/177, 382/180, 382/181, 382/209, 382/217, 382/270, 382/300, 356/121, 375/240.12, 375/240.26, 250/208.1, 250/201.9, 348/250
• US Class Current: 382/176
• CPC Class Codes: G06T 5/73 Deblurring; Sharpening
• CPC Class Codes: G06V 30/10 Character recognition
• CPC Class Codes: G06V 30/164 Noise filtering
• CPC Class Codes: G06V 30/186 by deriving mathematical or geometrical properties from the whole image
• CPC Class Codes: H04N 2101/00 Still video cameras
• CPC Class Codes: H04N 2201/0084 Digital still camera

Abstract

A system that extracts text from an image includes a capture device that captures the image having a low resolution. An image segmentation subsystem partitions the image into image segments. An image restoration subsystem generates a resolution-expanded image from the image segments and negates degradation effects of the low-resolution image by transforming the image segments from a first domain to a second domain and deconvolving the transformed image segments in the second domain to determine parameters of the low-resolution image. A text recognition subsystem transforms the restored image data into computer readable text data based on the determined parameters.

Claims

1. A system that extracts text from an image, comprising:

a capture device that captures the image having a low resolution;

an image segmentation subsystem that partitions the image into image segments;

an image restoration subsystem that generates a resolution-expanded image from the image segments and that negates degradation effects of the low-resolution image by transforming the image segments from a first domain to a second domain and deconvolving the transformed image segments in the second domain to determine parameters of the low-resolution image; and

a text recognition subsystem that transforms the restored image data into computer readable text data based on the determined parameters.

2. The system of claim 1, wherein the image segmentation subsystem identifies text from the image segments.

3. The system of claim 1, wherein the image restoration subsystem negates the degradation effects of the low-resolution image by negating distorting effects of defocus and CCD senor point spread function,

4. The system of claim 3, wherein the degradation of the low-resolution image is modeled by a defocus process and a low-resolution process.

5. The system of claim 1, wherein the capture device comprises a charge-coupled device.

6. The system of claim 1, wherein the image restoration subsystem includes a blind image restoration system.

7. The system of claim 6, wherein the blind image restoration system applies an iterative blind deconvolution (IBD) framework.

8. The system of claim 1, wherein the image restoration subsystem includes a Radon transform that transforms the image segments from the first domain to the second domain.

9. The system of claim 8, wherein the second domain is a Radon domain and the first domain is an image domain.

10. The system of claim 9, wherein the image restoration subsystem performs a de-averaging operation in the Radon domain to negate the degradation effects.

11. The system of claim 10, wherein the de-averaging operation is performed in a Radon domain.

12. The system of claim 8, wherein the image restoration subsystem includes an inverse Radon transform that transforms the image segments from the second domain to the first domain.

13. The system of claim 1, wherein the determined parameters of the low-resolution image are deconvolved by applying a Mutual Information Minimization (MIM)-based model.

14. The system of claim 13, wherein the image restoration subsystem performs a de-blurring operation to negate the degradation effects using the Mutual Information Minimization (MIM)-based model.

15. The system of claim 1, wherein the image segmentation subsystem identifies image segments that do not correspond to Whole characters by conforming with a holistic character recognition technique.

16. The system of claim 1, wherein the system restores images of non-cursive text.

17. A system that expands the resolution of a text image, comprising:

a text segmentation subsystem that partitions the text image into characters; and

a restoration subsystem comprising;

a first transform subsystem that transforms the partitioned text image from an image domain into a Radon domain;

an interpolator that linearly interpolates each column of the Radon domain image;

a first deconvolution subsystem that deconvolves each column to estimate first parameters of the text image;

an image expander that forms an expanded Radon domain image;

a second transform subsystem that transforms the expanded Radon domain image into the image domain;

an image slicer that slices the expanded image domain image to enforce a first constraint;

a third transform subsystem that transforms the sliced image from the image domain into the Radon domain;

a second deconvolution subsystem that estimates second parameters of the text image in the Radon domain;

a fourth transform subsystem that transforms the expanded Radon domain image including the second parameters into the image domain; and

an output that outputs the expanded image domain image as a restored text image.

18. The system of claim 17, wherein the first constraint is that the test image is bimodal.

19. The system of claim 17, wherein each column is deconvolved using estimated information in an assumed Gaussian channel.

20. The system of claim 17, wherein the expanded text image is stored, and wherein the second parameters are used as a prior estimate for a next text image.

21. The system of claim 17, wherein the system restores images of non-cursive text.