CODEWITHSUNDEEP
javaimageimage-processingcrop
Akashdeep Saluja
Akashdeep Saluja

Reputation: 3089

Detecting position of crop marks in a scanned document

I am working on a project in which user gives us a scanned copy of a form, and we need to identify various fields within the image.

I have a form like this: enter image description here

The form has four crop marks/ registration marks at four corners of the page. Now the user is supposed to fill the form and give back a scanned copy of his filled form. The scanned form that we get may have been rotated by some angle. For instance see the rotated form below: enter image description here

Now to extract the exact fields out of form, i.e. to extract any particular field like Title in the scanned form, we need to have exact coordinates but since our image has been rotated by unknown angle. We are unable to do so.

I read about registration marks and their use to align the page to a standard form. I tried searching for these registration marks in the image, but since it is rotated we may not find the position of the mark in the rotated image.

I tried searching for the issue and found some questions on SO which although gave some direction but couldn't help much.

This Question gave me reference to LEADTools SDK which has functions to perform this tasks. Unfortunately this SDK is not for JAVA and along with that it's proprietary and not free.

Is there any other open source tools for the same purpose.

Additionally I am open to suggestions about other methods used to align the form.

Upvotes: 2

Views: 2404

Answers (2)

Akashdeep Saluja
Akashdeep Saluja

Reputation: 3089

I successfully identified crop marks in the scanned document. I've described the approach on my blog here.

Upvotes: 0

Gabriel Archanjo
Gabriel Archanjo

Reputation: 4597

You could use the coordinates of the markers in the corner of the document. Using the coordinates of the corners of the lines, you could measure the rotation angle of the paper in order to compensate it.

The post below addresses a similar issue:
"Image Processing Edge Detection in Java"

Below an approach to detect the coordinates using Java and Marvin.

output (some noise cause of the JPEG compression):

enter image description here

source code:

import marvin.image.MarvinImage;
import marvin.io.MarvinImageIO;
import marvin.util.MarvinAttributes;
import static marvin.MarvinPluginCollection.*;

public class DocumentMarks {

public DocumentMarks(){

    MarvinImage image = MarvinImageIO.loadImage("./res/document.jpg");
    thresholding(image, 250);
    MarvinAttributes ret = moravec(image.clone(), image, 5, 100000);
    image = showCorners(image, ret, 4);
    MarvinImageIO.saveImage(image, "./res/document_out.jpg");
}

private static MarvinImage showCorners(MarvinImage image, MarvinAttributes attr, int rectSize){
    MarvinImage ret = image.clone();
    int[][] cornernessMap = (int[][]) attr.get("cornernessMap");
    int rsize=0;
    for(int x=0; x<cornernessMap.length; x++){
        for(int y=0; y<cornernessMap[0].length; y++){
            // Is it a corner?
            if(cornernessMap[x][y] > 0){
                rsize = Math.min(Math.min(Math.min(x, rectSize), Math.min(cornernessMap.length-x, rectSize)), Math.min(Math.min(y, rectSize), Math.min(cornernessMap[0].length-y, rectSize)));
                ret.fillRect(x, y, rsize, rsize, Color.red);
            }               
        }
    }

    return ret;
}

public static void main(String[] args) {
    new DocumentMarks();
    System.exit(0);
}
}

Upvotes: 3

Related Questions