Improve accuracy of tesseract ocr in android through preprocessing - java

The goal is to make on ocr app using tesseract, I didn't wanted to use tess-two as it works on older version of tesseract. So after a little research i was able to find this library which uses tesseract 4 and is a fork of tess-two. I am able to extract text from the image using this library. But my issue is Sometimes the same image gives 80% text and sometimes it gives 2% result. Here is the image :
and ocr result
, whereas the images having table format doesn't gives any result, only random letters.
I am new to ocr, pre-processing, can anyone help me for how can i improve the accuracy of image? I have read pre processing can be done, would that help and how it can be executed?
Thank you!

Related

How to detect image feature with java opencv?

i like to make feature detections using java android studio and in these past few day, i had been stuck on how to detect this kind of image using opencv.
Cause its my first time using openCV and the only document i can read is not really helpfull. I've already convert it to grayscale. The big questions mark is how can i make my apps to recognize the feature of this image.
so,for example i can make the label of this image as "abstractimage", so when i foto some of another image with the same pattern it will be detected as "abstractimage", but the detected image will be using on t-shirt, and random object with the same image based on this feature image. Can anyone help me?

Finding characteres in a game image using Java

I'm trying to recognize the characters of a game using Java, so I can process them later.
The images look like this:
http://i.imgur.com/GM0XPgO.jpg
I want to recognize, specially, the 3w in the left-bottom, the 214 and 0.13 in the center and the 1m in the right-bottom.
Anyone have any idea?
This question is related to computer vision and I would suggest using this java OCR library from here:
http://ocrsdk.com/documentation/code-samples/
This will help you recognize English text present on image. I'm pretty sure that it will give you major texts from image.
PS: For image processing, using matlab or python would be better as they have good computer vision libraries like "Computer vision Toolbox for Matlab" and "OpenCv for python". Here are some examples for you task using matlab: http://www.mathworks.com/help/vision/ref/ocr.html

Read Text from Image with Android

How to read text from image in android app.
Edited:
I want to detect text from image which i have captured from camera in android.
Is it possible or is there any library to read text from taken image.
This is not related to android. If you like to get text contained in image you will need
detect text position in image
perform OCR on found text
This task is not trivial and requires some computing resources. There are some OCR libraries around - like opencv, tesseract etc. I and others also develop pure java opensource solution:
http://sourceforge.net/projects/javaocr/

Android - Bitmap OCR

I get this problem:
The type java.awt.image.RenderedImage cannot be resolved. It is indirectly referenced from required .class files
I know it means that there is no reference to it in my build-path and I heard that java.awt.image is not in the Android SDK. So I am trying to figure out how to work around it.
This is my code:
Bitmap image = (Bitmap)data.getExtras().get("data");
String text = new OCR().recognizeEverything(image);
Obviously you can see that I am trying to use a OCR library. If it is impossible to get around this then can anyone show me a reference to maybe a "how to make an OCR program" or something along those lines. I pretty much have NO experience with images.
Thanks!
If your OCR library uses java.awt classes internally, you can't use it on Android. Porting it to use Android classes instead is likely to be non-trivial, especially if you have no experience. This library is being actively developed, and is reported to work on Android (I haven't personally used it). You might want to give it a try. Also, searching helps too: this is a fairly frequent question on SO, you might get some other ideas from previous answers.
We are developing pure java OCR library here:
http://sourceforge.net/projects/javaocr/
At the moment, there is some image processing stuff, invariant moment based recognition
and high performance binarisation. There are also demos schowcasing complete roundtrip for android ( gathering samples, performing recognizer training, performing reconigtion )
I already published 2 appluications based on it:
http://www.pribluda.de/android/charger/
http://www.pribluda.de/android/ocrcall/

Tracking multiple QR tags from a distance using webcam and Java

I'm currently trying to build an application in Java, which should be able to read QR tags from a couple of meters.
At the moment, I have JavaCV running, for capturing frames from my iSight on my MBP.
The problem is that it only supports a resolution of 640x480, which does not seem enough for ZXing to recognizing the QR tags.
I have been looking at a Logitech QuickCam Pro 9000, but I'm not entirely convinced, from my search, that it supports a high enough resolution for my project.
The application need to be platform agnostic, at least regarding OS X, Windows and Linux, which is the reason I'm doing it in Java.
My questions are:
Will the QucikCam pro record frames in 1080p as promised using the UVC drivers?
Is there a better camera?
Should I look at another framework than JavaCV and/or ZXing?
-Which?
thanks for the help
Resolution is not the issue; it decodes 320x240 regions on phones, as Barcode Scanner, easily. QR codes are especially simple to read. However if your QR code is in a tiny region of that 640x480 capture, like 50x30, yeah that's not going to be enough to decode reliably. I don't think any software could help you there.
I'd look at zooming in if you can.
Otherwise, I think you have some other issue; maybe the image data is in the wrong format somewhere

Categories

Resources