Reading Barcodes from an Image - II

• Download demo project - 59 Kb
• Download source - 75.4 Kb

Introduction

I recently ran across Benjamin Liedblad's article on reading barcode strings from images. The code was good, but not very useful in the real world. I thought my additions had a place here.

Background

Liedblad's barcode reader has two significant flaws:

• It only reads right-side up barcodes (as he mentioned in his "todo").
• It assumes that there is nothing other than the barcode in the image.

Frequently, we need to read a barcode from a page full of other text. This is what I set about to do.

Using the code

First of all, you should check out the original article to see how the barcode reading works. Much of the code is left unchanged.

Changes:

• Since the barcode we are looking for may not be at the beginning of the text line that we scan, we can't just scan by blocks of 9 'characters'. Instead, we walk along the scanned pattern one 'bar' at a time, testing it and the 8 bars after it to see if they form a character. If we find a real character, then we skip over the character pattern and begin again.
• Since there are text and other distractions on many pages, we need to cut the page up into sections and look for barcodes in each section. I've found that 50 is a good number of sections. While this seems like a lot, it's enough to ensure that we get the barcode we're looking for. Your mileage may vary. To this end, I added a startheight and endheight to ReadCode39(). Then we just need to calculate where each section begins and ends. This is accomplished easily in a for loop:

  for (int i=0; i < numscans; i++)
  {
      read = ReadCode39(bmp,i * (bmp.Height / numscans), 
             (i * (bmp.Height / numscans))+ (bmp.Height / numscans));
      [...]
  }

• We need to check all four page orientations, which is handled easily with .NET's Bitmap control:

  for (int i=0; i < 4; i++)
  {
      bmp.RotateFlip(System.Drawing.RotateFlipType.Rotate90FlipNone);
      VScanPageCode39(ref CodesRead, bmp,numscans);
  }

  You can, of course not use this if you know that all of your pages are the right way up.

• Finally, since we scan so many times and receive so many barcodes (including the backwards version of what we are looking for), we need to store all of the barcodes that we found. For this, I decided to use an ArrayList. The code currently returns all of the barcodes, but since most people supply some sort of pattern to their code (usually beginning and ending with astericks "*"), it's easy enough to pick out the right one.

Running the Demo

When you run the demo and load "SamplePage.jpg", first perform "Scan Barcode". This is essentially identical to Liedblad's original code. You will get "*JR7". Not very helpful. Then perform "Scan Page". Now you have a list of barcodes, one of which is the one you want (*surprise*)... and one is its mirror image (P4 9V*VK P). As you can see, it's pretty easy to pick out what you need.

History

• 5-15-05: Original.