amartin's profile - activity

Hello,

I'm very new to OpenCV. I'm trying to detect text regions containing headlines of a newspaper. I have read about OpenCV features for text detection:

http://docs.opencv.org/3.0-beta/modul...

But I'm not sure which steps to follow in order to reduce the image information in the newspaper and focus on the headlines.

Thank you very much.

Hello, I' m new in Opencv programing and image processing.

I 'm trying to segment an image containing the front page of a newspaper. I would like to retrieve only the head titles in order to apply OCR tesseract. I have already tried MSER blob detection, Stroke Width Transform and Robust Text detection with poor results:

http://web.stanford.edu/~hchen2/paper... http://research.microsoft.com/pubs/14...

I m using Opencv and Python. What would be a possible image processing pipeline for detecting text regions containing a font size of 20 or 25?

Thank you for your advice and help.

Hello I am new in OpenCV and Android, I'm running the tutorial examples, nevertheless the camera orientation is rotated, I have tried this code to fix it:

 mRgba = inputFrame.rgba();
 Mat mRgbaT = mRgba.t();
 Core.flip(mRgba.t(), mRgbaT, 1);
 Imgproc.resize(mRgbaT, mRgbaT, mRgba.size());
 return mRgbaT;

This works fine for a minute or so, frames are captured and represented correctly during one minute, but after that I get a SIGSev BpMemory fail and the application crashes.

Is there any workaround to change the camera orientation without having to do a flip and image resize?

Many thanks for your help and advice.

Hello, I'm new in OpenCV.

I'm trying to stitch different related images together with OpenCV's default stitcher, I have stitched images,taken in an horizontal manner, successfully.

I would like to stitch images captured, considering that the parts of the image that has to be taken are divided in cells of a grid, and the process of capturing happens like traversing a matrix in a row column fashion.

Is it possible to stitch the different images taken in a single and unique image? What approach is recommended in this case?

Thank you for your help and advice.