Wsrtjtyk

I am trying to read text from an image using tesseract, the picture is high qualit, so tesseract is able to read the text with 95% accuracy, which is OK for me at this point, however tesseract in unable to read the text which is highlighted (selected), Please refer the PIC

How can i read the text from selected area using tesseract, Is there any way to identify which word is highlighted in the image ?

Code used to read and convert the text from image as below.

tesseract::TessBaseAPI *myOCR = new tesseract::TessBaseAPI();



//cout << "hello" << endl;



if (myOCR->Init("C:\QTSoftware\IODriver\", "eng")) {



    SaveLineLog(brdInd, "Unable to initialize tesseract engine", __LINE__); 



    return RC_TESSERACT_ENG_FAILURE;



    //fprintf(stderr, "Could not initialize tesseract.n");



    //return RC_THREAD_FAILURE;

    //exit(1);

}



SaveLineLog(brdInd, "tesseract engine is UP :)", __LINE__);



//strcpy_s(fileName, "C:\TEMP\T481Logs\FrameOCR23_0.jpg");



FILE *pFile;



fopen_s(&pFile,fileName, "r");    // Open picture



PIX* pix;                         // Image format from `leptonica`



pix = pixReadStreamBmp(pFile);



fclose(pFile);



//Pix *pix = pixRead(fileName);



//Pix *pix = pixReadStreamBmp(fileName);



//cout << "Tesseract - Pix : " << pix << endl;



if (pix == NULL)

{

   SaveLineLog(brdInd, "Pix failure", __LINE__); 



   return RC_TESSERACT_PIX_FAILURE;

}



myOCR->SetImage(pix);



char* outText = myOCR->GetUTF8Text(); 

If all images have this format, you can first invert colors, then use Threshold binarization to get rid of this shadow.
It's better always to pre-process and provide clean image to the OCR engine, it will probably enhance the detection percent too.