System and method for smart contour completion

Patent metadata

Field	Value
Publication number	US-12394019-B2
Application number	US-202217799524-A
Country	US
Kind code	B2
Filing date	Aug 1, 2022
Priority date	Sep 2, 2021
Publication date	Aug 19, 2025
Grant date	Aug 19, 2025

Abstract

The present disclosure recites a system and a method to evaluate the scanned images of tables, identify at least one of the noises, errors and incomplete contours present therein (that make it difficult to extract cell location and it's contains) and process the said images of tables to remove the detected noises/errors from the images of the table and provide final images of table with complete contours. Therefore, said system disclosed herein is configured to take an image of table with incomplete contours (i.e. containing line gaps), as input and provides output in form of a processed image with all contours completed.

First claim

We claim:

1. A computer implemented method for smart contour completion, said method comprising:
converting an input image into grayscale image for segregating foreground region from background region by an adaptive threshold mechanism;
extracting all horizontal lines and all vertical line from said converted image, separately;
combining the extracted horizontal lines with the extracted vertical lines to define a table image with N cells;
removing noisy edge segments from said table image;
scanning boundary of each cell in the table image for identifying the cells with broken contours, wherein the cells with broken contour are identified as Region of Interest (ROI);
obtaining joins from said table image and defining a kernel size from said joins;
defining exterior boundary for said table image; and
dilating the broken contours, within defined boundary, in each ROI with the defined kernel size, in response to determining that the broken contour in the ROI is an endpoint and is not near the join, for contour completion.

2. The method of claim 1 , wherein removing noisy edge segments from said table image comprises:
identifying lines present at least inside the cell and outside the defined boundary of the table image; and
eliminating said lines from the table image, wherein removing the lines identified outside the defined boundary of the table image is optional.

3. The method of claim 1 , wherein the kernel size is defined by selecting a join with maximum height and maximum width from all the joins obtained.

4. The method of claim 1 , wherein the background region is represented by black colour and the foreground region is represented by plurality of vertical and horizontal lines, separately, in white colour.

5. The method of claim 1 , further comprises defining a suitable kernel size among different kernel sizes for perfect dilation for contour completion.

6. A system to provide smart contour completion, said system comprising:
an image conversion unit configured to convert an input image into grayscale image to segregate foreground region from background region by an adaptive threshold mechanism;
an image extractor operatively coupled to the image conversion unit and configured to extract all horizontal lines and all vertical line from said converted image, separately;
an adder configured to combine the extracted horizontal lines with the extracted vertical lines to define a table image with N cells; and
a processing unit operatively coupled to the adder, wherein the processing unit is configured to:
remove noisy edge segments from said table image;
scan boundary of each cell in the table image to identify the cells with broken contours, wherein the cells with broken contour is defined as Region of Interest (ROI);
obtain joins from said table image and define a kernel size from said joins;
define exterior boundary for said table image; and
dilate the broken contours, within defined boundary, for each ROI with the defined kernel size, in response to determining that the broken contour in the ROI is an endpoint and is not near the join, for contour completion.

7. The system of claim 6 , wherein to remove the noisy edge segments from said table image said processor is configured to:
identify lines present at least inside the cell and outside the defined boundary of the table image; and
eliminate said lines from the table image, wherein removing the lines identified outside the defined boundary of the table image is optional.

8. The system of claim 6 , wherein the processor is further configured to define the kernel size by selecting a join with maximum height and maximum width from all the joins obtained.

9. The system of claim 6 , wherein the image extractor is configured to segregate the background region by black colour and the foreground region by plurality of vertical and horizontal lines, separately, in white colour.

10. The system of claim 6 , wherein the processor is further configured to define a suitable kernel size among different kernel sizes for perfect dilation for contour completion.

Assignees

• L&T Technology Services Ltd

Inventors

• Mallick Triptesh
• Das Tarun Kumar
• Balaraman Mridul
• Singh Madhusudan

Classifications

• G06V10/46

  Descriptors for shape, contour or point-related descriptors, e.g. scale invariant feature transform [SIFT] or bags of words [BoW]; Salient regional features (colour feature extraction G06V10/56) · CPC title

• G06V10/30 Primary

  Noise filtering · CPC title

• G06V10/25

  Determination of region of interest [ROI] or a volume of interest [VOI] · CPC title

• G06T7/194

  involving foreground-background segmentation · CPC title

• G06V30/412

  Layout analysis of documents structured with printed lines or input boxes, e.g. business forms or tables · CPC title