Looking for support on adaptively cleaning up the background.

[deepstein]

Hello ImageMagick experts,

I need help in the image processing of an image with text + background. Currently i have written a script to read the text by adaptively eliminating the background. But it is not working on all images because of the variation in size, saturation, exposure, resolution and brightness of the image.

Current procedure:
1. Apply Bradley Roth for adaptive thresholding.
2. Using Image magick commands (-morphology close) to clean up specs/blobs from the background and in between text.
3. Tesseract to read the text.

Any advice/help in this regard is much appreciated and we are willing to pay for the support.

Thanks!

[fmw42]

Bradley Roth thresholding is likely similar to ImageMagick -lat or my script, localthresh (http://www.fmwconcepts.com/imagemagick/ ... /index.php). However, the mean statistics in ImageMagick are computed in a brute force way for each region; whereas in Bradley Roth, they are done using Integral Images. See user snibgo's page on Integral Images at http://im.snibgo.com/windmnsd.htm and http://im.snibgo.com/integim.htm

Post a few of your input image images if you want us to do some testing.