I am using Tesseract OCR (via pytesser) and PIL (Python Image Library) for automated test of an application.
I am checking that the displayed text is ok by making a screenshot and getting the text thanks to tesseract.
I had some issues in the beginning and it seems to work better since I have increased the size of the screenshot thanks to the bicubic interpolation of PIL.
Unfortunatelly, I still have some mistakes like confusion between '0' and 'O'. I can imagine that I will have other similar issues in the future.
I would like to know if there are some techniques to prepare an image in order to help the OCR. Any idea is welcomed.
Thanks in advance
Shameless plug and disclaimer: my company packages Tesseract for use in .NET
Tesseract is an OK OCR engine. It can miss a lot and gets readily confused by non-text. The best thing you can do for it is to make sure it gets text only. The next best thing is to give it something sanely binarized (adaptive or dynamic threshold to get there) or grayscale and let it try to do binarization.
Train tesseract to recognize your font
Make image extra clean and with enough free space around characters
Profit :)
Here are few real world examples.
First image is original image (croped power meter numbers)
Second image is slightly cleaned up image in GIMP, around 50% OCR accuracy in tesseract
Third image is completely cleaned image - 100% OCR recognized without any training!
Even under the best conditions OCR variants will sneak up on you. Your best option will be to design your tests to be aware of them.
For distinguishing between 0 and O, one simple solution is to choose a font that distinguishes between both (eg: 0 has a dash or dot in its middle). Would that be acceptable in your application?
Another solution is to apply a dictionary-based step after the character-by-character analysis of the text - feeding the recognized text into some form of spell-checker or validator to differentiate between difficult characters.
For instance, a round symbol followed by other numbers is most likely to be a zero, while the same symbol followed by letters is most likely to be a capital o. It's a trivial example, but it shows how context is necessary to make a more reliable OCR system.
Related
The simplest example of what I'm trying to recognize:
I use DOCUMENT_TEXT_DETECTION, but in the answer I get the hieroglyphics.
If I use Eng in the ImageContext parameter for the addAllLanguageHints method, then I have 111 in result. Better, but still bad.
Is there any way to indicate that the numbers are recognised or somehow improve the results?
Also, how is the setRepeatedField option in ImageContext is used? I could not find any examples of its use.
Thanks in advance.
Even if it doesn't work out of the box ... you'd need is to classify images using custom labels, when the default labels won't suffice. Cloud Auto ML Vision (select Vision from that blue drop-down menu) let's you train custom models, which can be used to recognize that font. And since the possible amount of shapes is quite limited with that 7-segment display, it shouldn't be too difficult to train it. If you'd get a calculator with a better display, it might also work better. The LCD above looks a little cheap, with those huge spaces and cut-off endings - but nevertheless, one can train it to read that.
I am trying to produce some high quality 8bpp bmp from pdf file with ghostscript. For that purpose, I use the bmp256 device.
So far, everything works well and is really fast, but ghostscript use halftoning to dither the image, leading to some uggly patterns when zooming on the picture :
I've managed to reduce their size by playing with the -dDITHERPPI flag, but this is still not satisfying enough. Those are too regular and are too easily spotted, even with little zoom.
Instead of using halftone, I would like to use some error diffusion algorithm, like the Floyd–Steinberg one. I found this algorithm is implemented on other devices, but they are all printer related devices, so I can't really use them.
Plus, I need to be as fast as possible when converting the PDF to 8bpp BMP, and the outputed pictures are very large: so converting it to 24 or 32bpp BMP in the first place to dither it later with another tool is excluded.
I already downloaded the source to try to implement it myself, but the project is really big and complex and I don't know how and where to start.
Is there any way to use some error diffusion algorithm with ghostscript without having to implement it myself ?
If no, is there a prefered way for extending ghostscript ? Any guideline ?
I mean posts per sentence, not per letter. Such a doctor's prescription handwriting which hard to read. Not just a normal handwriting.
In example :
I use a data mining or machine learning for doing a training from
paper handwrited.
User scanning a paper with hard to read writing.
The application doing an image processing.
And the output is some sentence from paper.
And what device to use? (Scanner or webcam)
I am newbie. If could i need some example in vb.net with emguCV/openCV and researches journals.
Any help would be appreciated.
Welcome to stack overflow! The answer to your question is twofold:
a. If you want to recognize handwriting that has already happened i.e. it is presented to you as an image you are in trouble. Computer Vision is still not good enough to provide you with reasonable accuracy.
b. If you have a chance to recognize handwriting “as it's happening” - you are in luck. Download, for example, a Gesture Search app from Android play store and you are in business.
The difference between the two scenarios is subtle but significant. In the second case you have an extra piece of information that makes handwriting recognition possible. This piece is timing of each stroke. In other words, instead of an image with handwriting you have a bunch of strokes that are all labeled with their time stamps. You can think about it as a sequence of lines and curves or as image segmentation - in any way this provides a big hint for the system. Additional help comes from the dictionary on your phone but this is typically used by any handwriting system.
Android of course has an open source library for stroke recognition (find more on your own). If you still want to go for recognizing images though, you have to first detect text (e.g. as a bounding box) and second use any of the existing engines to process detected regions. For text detection I can recommend MSER. But be careful trying to implement even text detection on your own - you are entering a world of pain here ;). Here is an article that can help.
As for learning how to recognize text from images on the Internet - this can be your plan B or C or Z when you master above mentioned stages. Don’t try to abuse learning methods and make them do hard work for you - you will hit a wall if you don’t understand what’s going on under the hood.
I'm trying to do the training process, but I don't understand even how to start. I would like to train for read it numbers. My images are from real world, so it didn't go so good with the reading process.
It says that I have to have a ".tif" image with the examples... is a single image of every number (in this case) or a image with a lot of different types of number (same font, though)?
And what about the makebox? The command didn't work here.
https://code.google.com/p/tesseract-ocr/wiki/TrainingTesseract3
Could someone explain me better, at least how to start?
I saw a few softwares that do this more quickly, but I tryied one (SunnyPage 1.8) but isn't free. Anyone know any free software that does this? Or a good tutorial?
Using Tesseract 3, Windows 8 (32bits).
It is important to patiently follow the training wiki google code project site. If needed multiple times. It is an open source library and is constantly evolving.
You will have to create a training image(tiff) with a lot of different types of numbers probably should have all the numbers you wish the engine to recognize.
Please consider posting the exact error message you got with make box.
I think Tesseract is the best free solution available. You have to keep working and seek help from community.
There is a very good post from Cédric here explaining the training process for Tesseract.
A good free OCR software is PDF OCR X which is also based on Tesseract. I tried to copy my notes from German which I had scanned at 1200dpi, and the results were commendable but not perfect. I found that this website - http://onlineocr.net - is a lot more accurate. If you are not registered, it allows a maximum of 4mb file size from most image formats (BMP, PNG, JPEG etc.) and PDF. It can output them as a Word file, an Excel file or an txt file.
Hope this helps.
Hey, i'm using a program called pngquant to convert 24 bit PNGs to 8-bit PNGs. Everything seems to work fine, and I don't notice any loss of quality for icons and other images that don't contain too much colors. Now when I feed it a PNG photo with zillions of colours, it produces a PNG8 where I can see some quality loss.
I'd like to determine that quality loss programmatically. I'd like to know when converting a PNG24 to PNG8 is safe or not. Sort of what webpagetest.org does -- they tell you that this specific image will be smaller in size if converted to PNG8 and will not loose quality.
Any ideas?
Thanks.
This sounds like a full-reference image quality assessment problem.
The simplest way to approach this is to try computing the PSNR between the PNG24 and PNG8 images. This is a measure of the difference between the two images. The higher the PSNR, the less different the images are. After using your color quantization software, check if the PSNR is above some threshold (you'll have to determine that empirically), and if it is, then the quantization was "safe".
PSNR has its down sides, namely the fact that it doesn't always correspond to the way the human visual system works (for example, it neglects the phenomenon of spatial and contrast masking). Another metric, SSIM, attempts to take care of that problem, but is slightly more difficult to compute (here is an OpenCV implementation, though). You can use SSIM instead of PSNR in the thresholding approach I described above.
Here's another thread which you might find useful.
Quite simple. If the image you are converting from PNG24 to PNG8 has more thant 256 colors, you gonna loose quality. Do I missed something?
For development of pngquant I use my own SSIM tool, since the OpenCV-based one didn't seem to support gamma correction nor alpha channel properly.
When you run pngquant -v it will output amount of error introduced as MSE=n (n is mean square error — 0 is perfect quality).
The latest version has --quality setting which lets you set minimum required quality. If it can't achieve it, it won't save the file.