Opencv C++ Thresholding not working after absdiff, Barcode Detection

1
votes

I am trying to detect barcode from an image using OpenCv, I am using the gradient method: http://www.pyimagesearch.com/2014/11/24/detecting-barcodes-images-python-opencv/ With help from the following question: How to find the location of red region in an image using MATLAB?

Code:

Mat imgco,img,imgx,imgy,thresh;
imgco = imread("/Users/a/Desktop/barcode_01.jpg");

//### Convert image to grayscale ###
cvtColor(imgco,img,CV_BGR2GRAY);
imshow("img", img);
waitKey(0);

//### Finding horizontal and vertical gradient ###

Sobel(img,imgx,CV_16S,1,0,-1,50,0,BORDER_DEFAULT);
imgx = abs(imgx);
imshow("img", imgx);
waitKey(0);

Sobel(img,imgy,CV_16S,0,1,-1,50,0,BORDER_DEFAULT);
imgy = abs(imgy);
imshow("img", imgy);
waitKey(0);

absdiff(imgx, imgy, img);
imshow("img", img);
waitKey(0);

//### Low pass filtering ###
GaussianBlur(img, img, Size(9,9), 0);
imshow("img", img);
waitKey(0);

//### Applying Threshold ###
threshold(img,thresh,225,255,CV_THRESH_BINARY);
imshow("img", thresh);
waitKey(0);

Now the program works correctly upto gaussian blur, Somehow trying to threshold the image gives a blank output. Can anyone suggest what could be the issue?

Gaussian Output:Threshold Output:

Thank You.

UPDATE: Applying the threshold operation before AbsDiff, Results in thresholding correctly. Hence there is something related to using AbsDiff, that is messing up with thresholding. Any help?

UPDATE 2: Comparing my output with the following: http://www.pyimagesearch.com/wp-content/uploads/2014/11/barcode_gradient_and_detection.jpg I realize that I am getting a grey image while the author is getting a crisp black bacground? Even when we are performing the same operations. Thank you.

1
c++opencvimage-processingbarcode-scannerthreshold
your sobel ops produce 16bit images, while the values in your threshold are for 8bit - berak
could you add the original image (w/o any filtering) ? - berak
Hello, Yes. The Original Image: pyimagesearch.com/wp-content/uploads/2014/11/barcode_01.jpg I tried to scale it to 8bit with: img.convertTo(img, CV_8U, 0.00390625); Still thresholding gives same output. - Aakash Thakkar
Also, Applying the threshold operation before AbsDiff, Results in thresholding correctly. Hence there is something related to using AbsDiff, that is messing up with thresholding. - Aakash Thakkar

1 Answers

0
votes

Okay the solution:

1) No need to use scaling in the gradient operations.

2) Use of convertScaleAbs method to convert the image to 8bit is necessary.

img.convertTo(img, CV_8U, 0.00390625);

Does not work. Solves the problem! :) Helped by the author of pyimage!