Image box filtering for optical character recognition

1. A method for box filtering comprising:

obtaining, by a computing device, a form image, the obtained from image comprising a plurality of boxes and image content in the plurality of boxes;

identifying, by the computing device, a region of the obtained form image comprising the plurality of boxes included in the obtained form image;

loading a set of box removal parameters, the box removal parameters comprising a box type, sliding window, and a width parameter;

detecting a plurality of vertical lines in the region of the obtained form image from a plurality of pixels in the region, the plurality of pixels comprising pixels representing the plurality of vertical lines, wherein each vertical line in the plurality of verticals lines is detected from a respective summation of pixels in the sliding window that comply with a color requirement;

detecting, by the computing device, the plurality of boxes in the region according to the box type, detected plurality of vertical lines and the width parameter;

extracting, by the computing device, image content from the plurality of boxes; and

generating a preprocessed form with the plurality of boxes removed, the preprocessed form including the image content extracted from the plurality of boxes.

2. The method of claim 1 , further comprising:

performing optical character recognition on the image content in the preprocessed form to generate a text recognized form;

extracting character content from the text recognized form to generate extracted character content; and

storing the extracted character content.

3. The method of claim 1 , wherein detecting the plurality of vertical lines comprises:

projecting, by the computing device, the plurality of pixels in the region onto a horizontal axis to create a horizontal axis projection;

identifying, by the computing device, a first plurality of peaks in the horizontal axis projection, wherein the plurality of vertical lines are detected using the first plurality of peaks.

4. The method of claim 3 , wherein detecting the plurality of boxes in the region comprises:

projecting the plurality of pixels on a vertical axis to create a vertical axis projection; and

identifying a second plurality of peaks in the vertical axis projection to detect a top and a bottom of each of the plurality of boxes.

5. The method of claim 4 , wherein detecting the plurality of boxes in the region comprises:

removing a sub-region of the region between adjacent peaks in the first plurality of peaks based on lacking the top and the bottom.

6. The method of claim 3 , wherein detecting the plurality of boxes comprises:

detecting a standard width of the plurality of boxes based on the plurality of peaks.

7. The method of claim 6 , wherein detecting the plurality of boxes comprises:

removing at least one peak from the first plurality of peaks that defines a boundary of a box that is less than the standard width by a threshold amount.

8. The method of claim 6 , wherein detecting the plurality of boxes comprises:

removing at least one peak from the first plurality of peaks that defines a boundary of a box that is greater than the standard width but not a multiple of the standard width.

9. The method of claim 1 , wherein extracting image content from the plurality of boxes comprises:

copying an image in each of the plurality of boxes within a defined margin.

10. A system for box filtering comprising:

a data repository for storing a form image;

a computer processor:

a box removal tool embodied as computer readable program code on a non-transitory computer readable medium, the computer readable program code executable to:

obtain the stored form image from the data repository;

load a set of box removal parameters, the box removal parameters comprising a box type, sliding window and a width parameter;

identify a region of the obtained form image comprising a plurality of boxes included in the obtained form image and image content in the plurality of boxes;

detect a plurality of vertical lines in the region of the obtained form image from a plurality of pixels in the region, the plurality of pixels comprising pixels that represent the plurality of vertical lines, each vertical line in the plurality of vertical lines detected from a respective summation of pixels in the sliding windowing complying with a color requirement;

detect the plurality of boxes in the region according to the box type, the detected plurality of vertical lines and the width parameter;

extract the image content from the plurality of boxes;

generate a preprocessed form with the plurality of boxes removed and that includes the image content extracted from the plurality of boxes.

11. The system of claim 10 , further comprising:

an optical character recognition (OCR) engine executable by the processor to perform OCR on the image content of the preprocessed form to generate a text recognized form;

a content extractor executable by the processor to:

extract character content from the text recognized form to generate extracted character content; and

store the extracted character content.

12. The system of claim 10 , further comprising a user interface executable by the processor to receive a plurality of box removal parameters.

13. The system of claim 12 , wherein the plurality of box removal parameters comprises a definition of the region, wherein the definition of the region comprises whitespace surrounding the plurality of boxes.

14. The system of claim 12 , wherein the plurality of box removal parameters comprises a definition of a margin within a box for extracting the image content.

15. A non-transitory computer readable medium for box filtering comprising computer readable program code for:

receiving, by a computing device, a form image;

loading a set of box removal parameters, the box removal parameters comprising a box type, sliding window and a width parameter;

identifying, by the computing device, a region of the received form image comprising a plurality of boxes included in the received form image, the received form image comprising the plurality of boxes and image content in the plurality of boxes;

detecting a plurality of vertical lines in the region of the received form image from a plurality of pixels in the region, the plurality of pixels comprising pixels that represent the plurality of vertical lines, wherein each vertical line in the plurality of vertical lines is detected from a respective summation of pixels in the sliding window that comply with a color requirement;

detecting the plurality of boxes in the region according to the box type, the detected plurality of vertical lines and the width parameter;

extracting image content from the plurality of boxes;

generating a preprocessed form with the plurality of boxes removed, the preprocessed form including the image content extracted from the plurality of boxes.

16. The non-transitory computer readable medium of claim 15 ,

wherein the non-transitory computer readable medium further comprises computer readable program code for:

performing optical character recognition (OCR) on the image content in the preprocessed form to generate a text recognized form;

extracting character content from the text recognized form to generate extracted character content; and

storing the extracted character content.

17. The non-transitory computer readable medium of claim 15 , wherein detecting the plurality of vertical lines comprises:

projecting, by the computing device, the plurality of pixels in the region onto a horizontal axis to create a horizontal axis projection;

identifying, by the computing device, a first plurality of peaks in the horizontal axis projection, wherein the plurality of vertical lines are detected using the first plurality of peaks.

18. The non-transitory computer readable medium of claim 17 , wherein detecting the plurality of boxes in the region comprises:

projecting the plurality of pixels on a vertical axis to create a vertical axis projection; and

identifying a second plurality of peaks in the vertical axis projection to detect a top and a bottom of each of the plurality of boxes.

19. The non-transitory computer readable medium of claim 18 , wherein detecting the plurality of boxes in the region comprises:

removing a sub-region of the region between adjacent peaks in the first plurality of peaks based on lacking the top and the bottom.

20. The non-transitory computer readable medium of claim 15 , wherein detecting the plurality of boxes comprises:

detecting a standard width of the plurality of boxes based on the plurality of vertical lines.