Hi James, Here I think more effort needs to be taken for getting better source images. In principle, there are two alternatives: - Get a good quality source image. Then you'll be able to handle it by means of relatively simple preprocessing. Maybe using ImageMagick. Probably you'll be able to use Tesseract. - Let any arbitrary image to get to your pipeline. Prepare to develop (or order from a 3rd party) complex image processor, full-fledged programming, etc.
If you choose to go with the first, I suggest the following to be improved to simplify further OCR: - Don't use JPEG. Because of that, there's massive bunch of compression artifacts in each of you images. Use lossless PNG instead. - Improve lighting. Too dark shots result in overwhelming noise. Either external or use flash. Beware of flares, though. Experiment in order to get best shots. - Try to hold camera evenly when shooting (fronto-parallel projection). Otherwise you'd need perspective correction as a preprocessing step. Or at least skew correction. - LCD display area to occupy as much as possible area of the image, centered. Otherwise you'd need background removal, ROI detection or devise heuristics for locating reference points in the image. If you fix all of the above, you'll probably be able to manage with the homemade ImageMagick scripts and Tesseract. You can send your sample images again, so that we can discuss what can be done further. There's a number of training attempts for LCD display fonts on the internet - look for them. They seem to address fonts similar to yours, but in the end you'd probably need to train yourself. Best regards, Dmitri Silaev www.CustomOCR.com On Thu, May 14, 2015 at 8:17 PM, James Okken <[email protected]> wrote: > Dmitri, > > thanks very much for your response. any help would be huge! > anything you suggest for LCD segments would be huge too! > > I've attached more of the original images. > > thanks > > On Thursday, May 14, 2015 at 3:41:14 AM UTC-4, Dmitri Silaev wrote: >> >> Hi James, >> >> I can suggest a number of steps regarding connected component analysis >> but it's better you'd show the original photo images. Probably there are >> easier ways to get the numbers from them. Be aware also that Tesseract >> might not be the best way to read LCD segment displays. It can work well >> for you, though; it depends on source image specifics. Attach several >> samples. >> >> Best regards, >> Dmitri Silaev >> www.CustomOCR.com >> >> >> >> >> >> On Wed, May 13, 2015 at 8:31 PM, James Okken <[email protected]> wrote: >> >>> hi everyone. >>> >>> can tesseract pull the numbers off this thermostat picture attached? >>> I've tried a number of things including making the photo a better quality, >>> to no avail. >>> >>> any help would be appreciated! thanks!! >>> >>> -- >>> You received this message because you are subscribed to the Google >>> Groups "tesseract-ocr" group. >>> To unsubscribe from this group and stop receiving emails from it, send >>> an email to [email protected]. >>> To post to this group, send email to [email protected]. >>> Visit this group at http://groups.google.com/group/tesseract-ocr. >>> To view this discussion on the web visit >>> https://groups.google.com/d/msgid/tesseract-ocr/0de8f0b4-dff2-44f0-bd91-bd0403e4d130%40googlegroups.com >>> <https://groups.google.com/d/msgid/tesseract-ocr/0de8f0b4-dff2-44f0-bd91-bd0403e4d130%40googlegroups.com?utm_medium=email&utm_source=footer> >>> . >>> For more options, visit https://groups.google.com/d/optout. >>> >> >> -- > You received this message because you are subscribed to the Google Groups > "tesseract-ocr" group. > To unsubscribe from this group and stop receiving emails from it, send an > email to [email protected]. > To post to this group, send email to [email protected]. > Visit this group at http://groups.google.com/group/tesseract-ocr. > To view this discussion on the web visit > https://groups.google.com/d/msgid/tesseract-ocr/41f963d3-e28b-47c2-a65e-c50ccef95530%40googlegroups.com > <https://groups.google.com/d/msgid/tesseract-ocr/41f963d3-e28b-47c2-a65e-c50ccef95530%40googlegroups.com?utm_medium=email&utm_source=footer> > . > > For more options, visit https://groups.google.com/d/optout. > -- You received this message because you are subscribed to the Google Groups "tesseract-ocr" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. To post to this group, send email to [email protected]. Visit this group at http://groups.google.com/group/tesseract-ocr. To view this discussion on the web visit https://groups.google.com/d/msgid/tesseract-ocr/CAKzLxFNr2La2jnhWfyT9T%3DTbogfgU-yGgOtYnHttDeqdMr_LhQ%40mail.gmail.com. For more options, visit https://groups.google.com/d/optout.
