Okay, if you’re running that high resolution I see that performance might be an issue…
But there’s a lot of memory copying in zbar-sharp, so you might want to do the croping in zbar-img.exe, and then maybe extend the program to loop through all files in the folder (to avoid spawning new processes).

Using the same 5 600dpi jpg images.

1. My program using command line zbar = 20 seconds.
2. My program using your library = 27 seconds.
3. My program using your library with cropping = 10 seconds.

Because we are feeding the documents through a document feeder the barcode is always aprox in the same spot in the images.

so 10 hours of processing will now be 5 hours hopefully

Also, if you performance in an issue why don’t you do it in C? Or at least the core functionality… Using ImageMagick for gray conversation might also yield better results.

But you’ll probably also experience a small performance improvement from using zbar-sharp… Because, unlike Linux, launching process on Windows is inherently slow.
(Do consider using latest sources from the master branch at github, they have fewer bugs).

1. try to read zbar installdir key from the regestry
2. if i can read it, show message that zbar isnt installed and launch zbar installer.
3. use that information to make a path from zbarimg.exe
4. have a way to select a folder on the UI.
5. for each file in the selected folder.
6. if its a jpg
7. make a new process
8. set the path to zbar that we worked out from the regestry
9. create the command line arguments including filename of the current file (including –raw command)
10. redirect both standard output and error output to textboxes on the form.

I guess i could have just used this…. oh well, i might rewrite it tomorrow at work to see if there is any speed increase.

I’m thinking that maybe zbar can consume color images… Maybe even RGB3 (24bit BMP), see:
http://sourceforge.net/apps/mediawiki/zbar/index.php?title=Supported_image_formats

Update: I had a look at the code and zbarimg uses ImageMagick for conversation to greyscale… Which is better than the greyscale conversation in zbar…

i took the latest master version from github from the link you have posted (13. January). in combination with the latest zbar (0.10).

The Bitmap scanner is the example you published at github as well. The core code is this:

System.Drawing.Image img = System.Drawing.Image.FromFile(file);

List symbols = scanner.Scan(img);
foreach (Symbol symbol in symbols)
Console.WriteLine(“\t” + symbol.ToString());

the file is the bitmap i send. The output is empty, so no symbols where found. The zbarimg.exe detects it and i also looked into that code (zbarimg.exe) and did not see any fancy stuff like configuration settings or similar things. so i assume the only different thing is the way how the image is converted. i was also not able to scan the following image too (an example image from zbar which also works with zbarimg which is an normal bar code): http://dl.dropbox.com/u/17637020/Just_right.png

besides the image conversion i do not see where the problem might be.

When, you say the bitmap scanner, is that the one in example/bmp-scanner ?
And what is the result? Does find anything, or what? bmp-scanner, doesn’t really print anything useful, it’s just an example…

i like this approach a lot and would also like to use it. I am facing some problems. I am using the bitmap scanner and was expecting similar results to what zbarimg is producing. But the result i am getting are rather frustrating. I use another component to capture a video stream and take an image and scan it afterwards. Uploaded an image which is working fine in a zbarimg and does not have a result with the bitmap scanner http://dl.dropbox.com/u/17637020/test.bmp.

am i missing something here?

In case anybody wonders the source is maintained here:
http://github.com/jopsen/zbar-sharp