OpenCV integration

[fullmetalcoder]

BTW: I think I found a bug in the documentation. In the format description, the QImage::Format_RGB32 is described as 0xffRRGGBB. However, it is 0xRRGGBBff. Or am I mistaken? But the above works!

I expect it depends on the endianness of your system...

[sfabel]

If it depended on the endianess of my system, it would be 0xBBGGRRff instead of 0xffRRGGBB, wouldn't it?

But since only the position of the alpha channel is changed I don't think it has anything to do with it.

Cheers,
Stephan

[pherthyl]

Hi Stephan,

I'm currently putting together a small project for a computer vision grad class I'm taking. Eventually it will be a head tracker, but right now it's just the general capturing framework that I will use. I did some benchmarking on different approaches to copying opencv data to a QImage, and ended up with the code posted here: http://www.qtcentre.org/forum/p-qima...ostcount7.html

I've also put together a simple framework for capturing and frame processing that you might find useful. I've attached the code for it. Just run qmake and make. Should work for most supported cameras.

It's multithreaded, so the capture thread spins getting frames as fast as possible and putting them into a buffer. Then the processing thread grabs frames out of that buffer and does whatever processing you want to do on the IplImage. Then the result gets loaded into a structure and passed through a chain of filters, ending with the widget that will contain the Image in the Qt GUI. Right now I do no processing and have no filters written yet, but the framework should work well for most computer vision tasks. I used a similar architecture for our eye tracking system and it works very well.

The display of the image uses a custom widget that paints it on the paintevent. I couldn't use OGL because my laptop video drivers are crap and it won't run. It should be about as fast as it can go without OGL though. Using Qt::WA_OpaquePaintEvent and Qt::WA_PaintOnScreen gives a nice speedup over the defaults.

[nooky59]

Hi,

Thanks for your code sample.

Just a question, is there any particular reason for not using the cvcam library bundled with opencv ? I have made a quick review of the code and I have noticed that you didn't used methods from cvcam.h

Your feedback on it could be very instructive.

[pherthyl]

I don't really have a good answer for you. I didn't use cvCam just because I haven't used it before, it seems to have less documentation (it's not even mentioned on the OpenCV wiki as far as I can tell), and highgui functions can do similar things.

However, if you are trying to build a real application that needs robust and full featured camera access, then I would advise against both highgui and cvcam. They are both more or less hacks that give you basic camera access but have all sorts of problems once you start using them for anything serious. For example, although highgui claims to be able to set things like camera resolution, framerate, and colour balance, in reality it doesn't actually work for most cameras.

If you have a good machine vision camera, use their provided SDK. It will always be miles better than the functions in opencv. Otherwise the way to go is to use the V4L api in Linux and the DirectX SDK in Windows. For example I have an Asus EeePC and the built in camera doesn't work properly with OpenCv at all, while it works fine in linux video apps and Windows video apps that don't use OpenCV.

[nooky59]

Thanks for your answer. Requirements for the projects are minimal for the moment so I don't think I need to deal with DirectShow or specific webcam driver.

For the eeePC, perhaps you should try with cvcam. I have tested succesfully with my logitech cam but this evening, I'm searching more informations on the benefits of cvcam against highgui and I've read this on a french forum :

else I know I was using capturefromcam (highgui.h) before with a logitech webcam and it was working like a charm, but when I switched to a wireless webcam and a video acquisition card, capturefromcam didn't worked anymore so I've used cvcam with the callback function and it work perfectly, so I think it support all webcams as long as drivers are installed

Perhaps it is worth a try ? Let me know if you get it working with eeePC both on Windows and Linux.

Here is a quick code sample :

#include <QtGui>
#include "mywidget.h"
#include "cvcam.h"
#include "cxcore.h"
#include "cxtypes.h"
 
void callback(IplImage* frame);
 
MyWidget::MyWidget(QWidget *parent) : QWidget(parent)
{
	int ncams = cvcamGetCamerasCount();
	cvcamSetProperty(0, CVCAM_PROP_ENABLE, CVCAMTRUE);
	cvcamSetProperty(0, CVCAM_PROP_RENDER, CVCAMTRUE);
 
	cvcamWindow myWin = (cvcamWindow) winId();
	cvcamSetProperty(0, CVCAM_PROP_WINDOW, &myWin);
 
	int width = 640;
	int height = 480;
	cvcamSetProperty(0, CVCAM_RNDWIDTH, &width);
	cvcamSetProperty(0, CVCAM_RNDHEIGHT, &height);
 
	cvcamSetProperty(0, CVCAM_PROP_CALLBACK, callback);
 
	cvcamInit();
	cvcamStart();
}
 
 
 
void callback(IplImage* frame)
{
// Do what you want with IplImage
}

#include <QtGui>
#include "mywidget.h"
#include "cvcam.h"
#include "cxcore.h"
#include "cxtypes.h"

void callback(IplImage* frame);

MyWidget::MyWidget(QWidget *parent) : QWidget(parent)
{
	int ncams = cvcamGetCamerasCount();
	cvcamSetProperty(0, CVCAM_PROP_ENABLE, CVCAMTRUE);
	cvcamSetProperty(0, CVCAM_PROP_RENDER, CVCAMTRUE);
	
	cvcamWindow myWin = (cvcamWindow) winId();
	cvcamSetProperty(0, CVCAM_PROP_WINDOW, &myWin);

	int width = 640;
	int height = 480;
	cvcamSetProperty(0, CVCAM_RNDWIDTH, &width);
	cvcamSetProperty(0, CVCAM_RNDHEIGHT, &height);

	cvcamSetProperty(0, CVCAM_PROP_CALLBACK, callback);

	cvcamInit();
	cvcamStart();
}



void callback(IplImage* frame)
{
// Do what you want with IplImage
}

To copy to clipboard, switch view to plain text mode 

I haven't bundled the termination code but you will found it in the RTF doc bundled with OpenCV. As there is some mistakes in the first code sample from the doc, I think the above code should help you (even if the remaining of the RTF is useful to understand and perhaps a litte more accurate ;o))

[patrik08]

I put this QImage IplImageToQImage here on this thread if people like to search...
I tested on QT4.4 and run ok....

from QPainter p(&gd); forward is only a date print...
I write & copy this piece to a label :
Animated Portable Network Graphics APNG , which run on Firefox 3 or opera...
http://www.qt-apps.org/content/show....?content=82221

#ifdef OPCAMENABLE
 
#include "cv.h"
#include "highgui.h"
#include <stdio.h>
#include <ctype.h>
 
static QImage IplImageToQImage(const IplImage  *iplImage, uchar **data , bool mirroRimage = true )
{
  uchar *qImageBuffer = NULL;
  int    width        = iplImage->width;
 
  /*
   * Note here that OpenCV image is stored so that each lined is 32-bits aligned thus
   * explaining the necessity to "skip" the few last bytes of each line of OpenCV image buffer.
   */
  int widthStep = iplImage->widthStep;
  int height    = iplImage->height;
 
  switch (iplImage->depth)
  {
    case IPL_DEPTH_8U:
      if (iplImage->nChannels == 1)
      {
        /* OpenCV image is stored with one byte grey pixel. We convert it to an 8 bit depth QImage. */
        qImageBuffer = (uchar *)malloc(width * height * sizeof(uchar));
        uchar *QImagePtr = qImageBuffer;
        const uchar *iplImagePtr = (const uchar *)iplImage->imageData;
        for (int y = 0; y < height; ++y)
        {
          // Copy line by line
          memcpy(QImagePtr, iplImagePtr, width);
          QImagePtr += width;
          iplImagePtr += widthStep;
        }
      }
      else if (iplImage->nChannels == 3)
      {
        /* OpenCV image is stored with 3 byte color pixels (3 channels). We convert it to a 32 bit depth QImage. */
        qImageBuffer = (uchar *)malloc(width * height * 4 * sizeof(uchar));
        uchar *QImagePtr = qImageBuffer;
        const uchar *iplImagePtr = (const uchar *)iplImage->imageData;
 
        for (int y = 0; y < height; ++y)
        {
          for (int x = 0; x < width; ++x)
          {
            // We cannot help but copy manually.
            QImagePtr[0] = iplImagePtr[0];
            QImagePtr[1] = iplImagePtr[1];
            QImagePtr[2] = iplImagePtr[2];
            QImagePtr[3] = 0;
 
            QImagePtr += 4;
            iplImagePtr += 3;
          }
          iplImagePtr += widthStep - 3 * width;
        }
      }
      else
        qDebug("IplImageToQImage: image format is not supported : depth=8U and %d channels\n", iplImage->nChannels);
 
      break;
 
    case IPL_DEPTH_16U:
      if (iplImage->nChannels == 1)
      {
        /* OpenCV image is stored with 2 bytes grey pixel. We convert it to an 8 bit depth QImage. */
        qImageBuffer = (uchar *)malloc(width * height * sizeof(uchar));
        uchar *QImagePtr = qImageBuffer;
        const uint16_t *iplImagePtr = (const uint16_t *)iplImage->imageData;
 
        for (int y = 0; y < height; ++y)
        {
          for (int x = 0; x < width; ++x)
            *QImagePtr++ = ((*iplImagePtr++) >> 8); // We take only the highest part of the 16 bit value. It is similar to dividing by 256.
          iplImagePtr += widthStep / sizeof(uint16_t) - width;
        }
      }
      else
        qDebug("IplImageToQImage: image format is not supported : depth=16U and %d channels\n", iplImage->nChannels);
 
      break;
 
    case IPL_DEPTH_32F:
      if (iplImage->nChannels == 1)
      {
        /* OpenCV image is stored with float (4 bytes) grey pixel. We convert it to an 8 bit depth QImage. */
        qImageBuffer = (uchar *)malloc(width * height * sizeof(uchar));
        uchar *QImagePtr = qImageBuffer;
        const float *iplImagePtr = (const float *)iplImage->imageData;
 
        for (int y = 0; y < height; ++y)
        {
          for (int x = 0; x < width; ++x)
            *QImagePtr++ = (uchar)(255 * ((*iplImagePtr++)));
          iplImagePtr += widthStep / sizeof(float) - width;
        }
      }
      else
        qDebug("IplImageToQImage: image format is not supported : depth=32F and %d channels\n", iplImage->nChannels);
 
      break;
 
    case IPL_DEPTH_64F:
      if (iplImage->nChannels == 1)
      {
        /* OpenCV image is stored with double (8 bytes) grey pixel. We convert it to an 8 bit depth QImage. */
        qImageBuffer = (uchar *) malloc(width * height * sizeof(uchar));
        uchar *QImagePtr = qImageBuffer;
        const double *iplImagePtr = (const double *) iplImage->imageData;
 
        for (int y = 0; y < height; ++y)
        {
          for (int x = 0; x < width; ++x)
            *QImagePtr++ = (uchar)(255 * ((*iplImagePtr++)));
          iplImagePtr += widthStep / sizeof(double) - width;
        }
      }
      else
        qDebug("IplImageToQImage: image format is not supported : depth=64F and %d channels\n", iplImage->nChannels);
 
      break;
 
    default:
      qDebug("IplImageToQImage: image format is not supported : depth=%d and %d channels\n", iplImage->depth, iplImage->nChannels);
  }
 
  QImage *qImage;
  if (iplImage->nChannels == 1)
  {
    QVector<QRgb> colorTable;
    for (int i = 0; i < 256; i++)
      colorTable.push_back(qRgb(i, i, i));
 
    qImage = new QImage(qImageBuffer, width, height, QImage::Format_Indexed8);
    qImage->setColorTable(colorTable);
  }
  else
    qImage = new QImage(qImageBuffer, width, height, QImage::Format_RGB32);
    QImage gd0 = qImage->mirrored(false,mirroRimage);
    *data = qImageBuffer;
    QColor textColor = Qt::black;
    QColor fillrectcolor = Qt::red;
    QColor shapepicture = Qt::white;
    QImage gd = gd0.scaledToWidth(350);
    QDateTime now = QDateTime::currentDateTime ();
    QString selectionText = now.toString("dd.MM.yyyy hh:mm:ss");
    QPainter p(&gd);
    p.setRenderHint(QPainter::Antialiasing, true);
 
    QFontMetrics fm( qApp->font() );
    int stringWidth = fm.width(selectionText);
    int stringHeight = fm.ascent();
    const int sx = gd.width() - stringWidth - 5;
    QPen pen;
    pen.setStyle( Qt::SolidLine );
    pen.setWidth( 2 );
    pen.setColor( textColor );
    p.setPen( pen);
    p.drawText(QPointF(sx - 1 ,gd.height() - 2 - stringHeight - 1),selectionText);
    pen.setColor( fillrectcolor );
    p.setPen( pen);
    p.drawText(QPointF(sx,gd.height() - 2 - stringHeight),selectionText);
 
  return gd;
}
 
 
 
#endif

#ifdef OPCAMENABLE

#include "cv.h"
#include "highgui.h"
#include <stdio.h>
#include <ctype.h>

static QImage IplImageToQImage(const IplImage  *iplImage, uchar **data , bool mirroRimage = true )
{
  uchar *qImageBuffer = NULL;
  int    width        = iplImage->width;

  /*
   * Note here that OpenCV image is stored so that each lined is 32-bits aligned thus
   * explaining the necessity to "skip" the few last bytes of each line of OpenCV image buffer.
   */
  int widthStep = iplImage->widthStep;
  int height    = iplImage->height;

  switch (iplImage->depth)
  {
    case IPL_DEPTH_8U:
      if (iplImage->nChannels == 1)
      {
        /* OpenCV image is stored with one byte grey pixel. We convert it to an 8 bit depth QImage. */
        qImageBuffer = (uchar *)malloc(width * height * sizeof(uchar));
        uchar *QImagePtr = qImageBuffer;
        const uchar *iplImagePtr = (const uchar *)iplImage->imageData;
        for (int y = 0; y < height; ++y)
        {
          // Copy line by line
          memcpy(QImagePtr, iplImagePtr, width);
          QImagePtr += width;
          iplImagePtr += widthStep;
        }
      }
      else if (iplImage->nChannels == 3)
      {
        /* OpenCV image is stored with 3 byte color pixels (3 channels). We convert it to a 32 bit depth QImage. */
        qImageBuffer = (uchar *)malloc(width * height * 4 * sizeof(uchar));
        uchar *QImagePtr = qImageBuffer;
        const uchar *iplImagePtr = (const uchar *)iplImage->imageData;

        for (int y = 0; y < height; ++y)
        {
          for (int x = 0; x < width; ++x)
          {
            // We cannot help but copy manually.
            QImagePtr[0] = iplImagePtr[0];
            QImagePtr[1] = iplImagePtr[1];
            QImagePtr[2] = iplImagePtr[2];
            QImagePtr[3] = 0;

            QImagePtr += 4;
            iplImagePtr += 3;
          }
          iplImagePtr += widthStep - 3 * width;
        }
      }
      else
        qDebug("IplImageToQImage: image format is not supported : depth=8U and %d channels\n", iplImage->nChannels);

      break;

    case IPL_DEPTH_16U:
      if (iplImage->nChannels == 1)
      {
        /* OpenCV image is stored with 2 bytes grey pixel. We convert it to an 8 bit depth QImage. */
        qImageBuffer = (uchar *)malloc(width * height * sizeof(uchar));
        uchar *QImagePtr = qImageBuffer;
        const uint16_t *iplImagePtr = (const uint16_t *)iplImage->imageData;

        for (int y = 0; y < height; ++y)
        {
          for (int x = 0; x < width; ++x)
            *QImagePtr++ = ((*iplImagePtr++) >> 8); // We take only the highest part of the 16 bit value. It is similar to dividing by 256.
          iplImagePtr += widthStep / sizeof(uint16_t) - width;
        }
      }
      else
        qDebug("IplImageToQImage: image format is not supported : depth=16U and %d channels\n", iplImage->nChannels);

      break;

    case IPL_DEPTH_32F:
      if (iplImage->nChannels == 1)
      {
        /* OpenCV image is stored with float (4 bytes) grey pixel. We convert it to an 8 bit depth QImage. */
        qImageBuffer = (uchar *)malloc(width * height * sizeof(uchar));
        uchar *QImagePtr = qImageBuffer;
        const float *iplImagePtr = (const float *)iplImage->imageData;

        for (int y = 0; y < height; ++y)
        {
          for (int x = 0; x < width; ++x)
            *QImagePtr++ = (uchar)(255 * ((*iplImagePtr++)));
          iplImagePtr += widthStep / sizeof(float) - width;
        }
      }
      else
        qDebug("IplImageToQImage: image format is not supported : depth=32F and %d channels\n", iplImage->nChannels);

      break;

    case IPL_DEPTH_64F:
      if (iplImage->nChannels == 1)
      {
        /* OpenCV image is stored with double (8 bytes) grey pixel. We convert it to an 8 bit depth QImage. */
        qImageBuffer = (uchar *) malloc(width * height * sizeof(uchar));
        uchar *QImagePtr = qImageBuffer;
        const double *iplImagePtr = (const double *) iplImage->imageData;

        for (int y = 0; y < height; ++y)
        {
          for (int x = 0; x < width; ++x)
            *QImagePtr++ = (uchar)(255 * ((*iplImagePtr++)));
          iplImagePtr += widthStep / sizeof(double) - width;
        }
      }
      else
        qDebug("IplImageToQImage: image format is not supported : depth=64F and %d channels\n", iplImage->nChannels);

      break;

    default:
      qDebug("IplImageToQImage: image format is not supported : depth=%d and %d channels\n", iplImage->depth, iplImage->nChannels);
  }

  QImage *qImage;
  if (iplImage->nChannels == 1)
  {
    QVector<QRgb> colorTable;
    for (int i = 0; i < 256; i++)
      colorTable.push_back(qRgb(i, i, i));

    qImage = new QImage(qImageBuffer, width, height, QImage::Format_Indexed8);
    qImage->setColorTable(colorTable);
  }
  else
    qImage = new QImage(qImageBuffer, width, height, QImage::Format_RGB32);
    QImage gd0 = qImage->mirrored(false,mirroRimage);
    *data = qImageBuffer;
    QColor textColor = Qt::black;
    QColor fillrectcolor = Qt::red;
    QColor shapepicture = Qt::white;
    QImage gd = gd0.scaledToWidth(350);
    QDateTime now = QDateTime::currentDateTime ();
    QString selectionText = now.toString("dd.MM.yyyy hh:mm:ss");
    QPainter p(&gd);
    p.setRenderHint(QPainter::Antialiasing, true);
  
    QFontMetrics fm( qApp->font() );
    int stringWidth = fm.width(selectionText);
    int stringHeight = fm.ascent();
    const int sx = gd.width() - stringWidth - 5;
    QPen pen;
    pen.setStyle( Qt::SolidLine );
    pen.setWidth( 2 );
    pen.setColor( textColor );
    p.setPen( pen);
    p.drawText(QPointF(sx - 1 ,gd.height() - 2 - stringHeight - 1),selectionText);
    pen.setColor( fillrectcolor );
    p.setPen( pen);
    p.drawText(QPointF(sx,gd.height() - 2 - stringHeight),selectionText);
  
  return gd;
}



#endif

To copy to clipboard, switch view to plain text mode 

[patelmiteshn]

Hello Forum,

I'm new to both QT and OpenCV. I've written couple of program in OpenCV and works well and done the same with QT so botht the tools are working well.

I'm trying to integrate the code posted by "pherthyl" to stream the video data caputured by the webcam in opencv. The application is really well written (for me as I can understand it). When i try and build it says that cannot find "opencv/highgui.h"

I've installed opencv in /opt/opencv and all the files are viz. highgui.h, cxcore.h, cv.h are located in /opt/opencv/include/opencv.

I tried and compile the project by changing the path of highgui.h in capturethread.h from opencv/highgui.h to opt/opencv/include/opencv/highgui.h but still it gives the same error that it couldn't locate the files.

Is there any specific changes that I need to make in the MAKEFILE which will specify the location of these files and link it.

Thanks in advance.

Regards,
Mitesh

[kosirm]

Hi Stephan,

I'm currently putting together a small project for a computer vision grad class I'm taking. Eventually it will be a head tracker, but right now it's just the general capturing framework that I will use. I did some benchmarking on different approaches to copying opencv data to a QImage, and ended up with the code posted here: http://www.qtcentre.org/forum/p-qima...ostcount7.html

I've also put together a simple framework for capturing and frame processing that you might find useful. I've attached the code for it. Just run qmake and make. Should work for most supported cameras.

It's multithreaded, so the capture thread spins getting frames as fast as possible and putting them into a buffer. Then the processing thread grabs frames out of that buffer and does whatever processing you want to do on the IplImage. Then the result gets loaded into a structure and passed through a chain of filters, ending with the widget that will contain the Image in the Qt GUI. Right now I do no processing and have no filters written yet, but the framework should work well for most computer vision tasks. I used a similar architecture for our eye tracking system and it works very well.

The display of the image uses a custom widget that paints it on the paintevent. I couldn't use OGL because my laptop video drivers are crap and it won't run. It should be about as fast as it can go without OGL though. Using Qt::WA_OpaquePaintEvent and Qt::WA_PaintOnScreen gives a nice speedup over the defaults.

Thanks for code!
I'm running example and capture frame size isn't changing, it seems to me that cvSetCaptureProperty(capture, CV_CAP_PROP_FRAME_WIDTH, 640); does nothing. Or have I missed something?

if(size == Size640) {
cvSetCaptureProperty(capture, CV_CAP_PROP_FRAME_WIDTH, 640);
cvSetCaptureProperty(capture, CV_CAP_PROP_FRAME_HEIGHT, 480);
qDebug() << "Setting 640x480:CV_CAP_PROP_FRAME_WIDTH:" << cvGetCaptureProperty(capture, CV_CAP_PROP_FRAME_WIDTH);
qDebug() << "Setting 640x480CV_CAP_PROP_FRAME_HEIGHT:" << cvGetCaptureProperty(capture, CV_CAP_PROP_FRAME_HEIGHT);

}

returns:
Setting 640x480:CV_CAP_PROP_FRAME_WIDTH: 320
Setting 640x480CV_CAP_PROP_FRAME_HEIGHT: 240

Any ideas how could I change frame width/height?
Best regards,
Milan

[pherthyl]

Thanks for code!
I'm running example and capture frame size isn't changing, it seems to me that cvSetCaptureProperty(capture, CV_CAP_PROP_FRAME_WIDTH, 640); does nothing. Or have I missed something?

if(size == Size640) {
cvSetCaptureProperty(capture, CV_CAP_PROP_FRAME_WIDTH, 640);
cvSetCaptureProperty(capture, CV_CAP_PROP_FRAME_HEIGHT, 480);
qDebug() << "Setting 640x480:CV_CAP_PROP_FRAME_WIDTH:" << cvGetCaptureProperty(capture, CV_CAP_PROP_FRAME_WIDTH);
qDebug() << "Setting 640x480CV_CAP_PROP_FRAME_HEIGHT:" << cvGetCaptureProperty(capture, CV_CAP_PROP_FRAME_HEIGHT);

}

returns:
Setting 640x480:CV_CAP_PROP_FRAME_WIDTH: 320
Setting 640x480CV_CAP_PROP_FRAME_HEIGHT: 240

Any ideas how could I change frame width/height?
Best regards,
Milan

No. This is what I was talking about when I said the highgui functions suck

On some cameras those functions will work, but on most they won't. If you want to control camera properties like capture size and others you'll have to use another image acquisition library or try cvcam as someone else suggested. I never had much luck with it. It'll do basic capture, but that's it.

[kosirm]

thanks...

best regards, Milan

[switch]

Sorry to re-awaken an old topic but i am having some difficulties converting an opencv Monochrome image into a Qimage. The image is a result of calling the opencv function
void cvInRangeS( const CvArr* src, CvScalar lower, CvScalar upper, CvArr* dst );

When i pass this image to the function IplImageToQImage it does convert the image into qimage however the image is 8 times smaller and duplicated. The results can be seen below

Opencv Image


Convertion to Qimage results



Does anybody know why this has happened? And any ideas how to correct the error?

[pherthyl]

Not sure why you're getting that. Have you tried looking at the dst image before you convert it to a QImage (using the highgui showimg function)? Make sure the dst image is the correct bit depth (8u or 8s).

[eumesmo]

Hello pherthyl,


Thanks for the framework, it's very good, but I found a problem running in my hardware. I tested the framework in a Hp notebook, with the webcam and with a external camera and worked fine. When a tried the framework in another notebook (a LG note) with the integrate webcam, a get an deadlock every time I stop and try a play again.

The console output: (this problem in this notebook occurs every time, it's not a random problem...)


Listing capture properties...
CV_CAP_PROP_FRAME_WIDTH
CV_CAP_PROP_FRAME_HEIGHT
CV_CAP_PROP__FPS 0
CV_CAP_PROP_FOURCC 0
CV_CAP_PROP_BRIGHTNESS 0
CV_CAP_PROP_CONTRAST 0
CV_CAP_PROP_SATURATION 0
CV_CAP_PROP_HUE 0
Done
Setting 640x480 1
Attempting to set frame rate...
Error: 0
Starting to track
About to start the capture thread
Started the capture thread
Stop capture requested.
Waiting on capture start...
Listing capture properties...
CV_CAP_PROP_FRAME_WIDTH
CV_CAP_PROP_FRAME_HEIGHT
CV_CAP_PROP__FPS 0
CV_CAP_PROP_FOURCC 0
CV_CAP_PROP_BRIGHTNESS 0
CV_CAP_PROP_CONTRAST 0
CV_CAP_PROP_SATURATION 0
CV_CAP_PROP_HUE 0
Done
Setting 640x480 1
Attempting to set frame rate...
Error: 0
Starting to track
About to start the capture thread
Started the capture thread
QMutex::lock Deadlock detected in thread 3916



Well, another test that I did, it's to start a HeadTracker object when a button is pressed in the main window. Well, it works fine until I request a re-size. After a re-size the program crashes and close.


It's just a report, but any help it's very welcome, I will have a better lock in the code to see if I can improve anything, because I think is a very good framework and really fast as well

cheers!

[mounte]

I know this thread is quite old, but still it is a very good wrapper to get images from webcam or whatnot.

If for some reason you dont need some fancy 32 bit format you could change the updatePixmap function to only use one memcpy like this:

QTime t;
    t.start();
    //qDebug() << "Copying data";
    bool start = false;
    // check if the frame dimensions have changed
    if(frame->width != imageWidth || frame->height != imageHeight) {
        if(imageData) {
            delete[] imageData;
        }
        start = true;
        imageWidth = frame->width;
        imageHeight = frame->height;
        emit(frameSizeChanged(imageWidth, imageHeight));
        imageData = new unsigned char[3*imageWidth*imageHeight];
    }
 
    int pixels = imageWidth * imageHeight;
    uchar* src = (uchar*)(frame->imageData);
 
    memcpy(imageData, src, pixels*3);
    if(!start) {
        ++frames;
        time += t.elapsed();
    }

QTime t;
    t.start();
    //qDebug() << "Copying data";
    bool start = false;
    // check if the frame dimensions have changed
    if(frame->width != imageWidth || frame->height != imageHeight) {
        if(imageData) {
            delete[] imageData;
        }
        start = true;
        imageWidth = frame->width;
        imageHeight = frame->height;
        emit(frameSizeChanged(imageWidth, imageHeight));
        imageData = new unsigned char[3*imageWidth*imageHeight];
    }

    int pixels = imageWidth * imageHeight;
    uchar* src = (uchar*)(frame->imageData);

    memcpy(imageData, src, pixels*3);
    if(!start) {
        ++frames;
        time += t.elapsed();
    }

To copy to clipboard, switch view to plain text mode 

Then you have to change the paintEvent method to create the tImg with RGB888 format like this:

QImage tImg(imageData, imageWidth, imageHeight, QImage::Format_RGB888);

QImage tImg(imageData, imageWidth, imageHeight, QImage::Format_RGB888);

To copy to clipboard, switch view to plain text mode 

You might run into problem of RGB <--> BGR conversion, just modify the painter.drawImage call to

painter.drawImage(QPoint(0,0), tImg.rgbSwapped());

painter.drawImage(QPoint(0,0), tImg.rgbSwapped());

To copy to clipboard, switch view to plain text mode 

[umanga]

Seems the archive is corrupted.Could you please send me this ? my email is

umanga dot pdn at gmail dot com

thanks

[josiahbryan]

I thought the archive was corrupted too - but its actually double-gzipped, then tarred. So to open the archive, I had to gunzip qtopencv.tar.gz, then "cat qtopencv.tar | gunzip > qtopencv2.tar", then I could do "tar xvf qtopencv.tar" to extract the files. To save everyone that work, I posted a zipped version of the archive at: http://www.jdbryanphotography.com/do...s/qtopencv.zip. Cheers!

[szworker]

Can we do like this? Mine is very simple. It works ( I think ) can you guys help check this?

IplImage * img = cvLoadImage( "C:\\Users\\Pictures\\dodo19.jpg");
cvCvtColor(img,img,CV_BGR2RGB);
m_i = QImage((unsigned char *)img->imageDataOrigin,img->width,img->height,QImage::Format_RGB888);
 imageLabel->setPixmap(QPixmap::fromImage(m_i,0));

IplImage * img = cvLoadImage( "C:\\Users\\Pictures\\dodo19.jpg");
cvCvtColor(img,img,CV_BGR2RGB);
m_i = QImage((unsigned char *)img->imageDataOrigin,img->width,img->height,QImage::Format_RGB888);
 imageLabel->setPixmap(QPixmap::fromImage(m_i,0));

To copy to clipboard, switch view to plain text mode 

The picture looks good in the Qt Dialog. Please check if doing like this does really work fine.

[srsr]

If it can help here is a code-snippet for a custom Qt plug-in using OpenCV. I use "neutral" unsigned char* between Qt and OpenCV. On return unsigned char* converted to QImage is displayed in a custom Widget. This PlugIn is used reall-time on video stream ( on PC.. sorry).

// code
#include <QtGui>
#include <math.h>
#include <stdlib.h>
#include "OpenCVwarpFilter.h"
//
#include "cv.h"
#include "cxcore.h"
 
#pragma message("automatic link to OpenCV libs")
#pragma comment(lib,"cv210.lib")
#pragma comment(lib,"cxcore210.lib")
 
struct OpenCVStruct
{
int             iWidth;
int             iHeight;
IplImage        *m_pImg;        // OpenCV Images
IplImage        *m_pOutputImg;
CvMat           *m_pMapMatrix;  // OpenCV Matrix
CvPoint2D32f     src_quad[4];
CvPoint2D32f     dst_quad[4];
};
 
QString OpenCVwarpFilterPlugin::filterName() const
{
    return "piWarp_OpenCV";
}
 
 
unsigned long OpenCVwarpFilterPlugin::filterParamsSize() const
{
    return sizeof(OpenCVStruct);
}
 
void OpenCVwarpFilterPlugin::filterInit(void* piParams)
{
	OpenCVStruct* pOpenCV = (OpenCVStruct*)piParams;
	pOpenCV->m_pImg		    = NULL;
	pOpenCV->m_pOutputImg	= NULL;
	pOpenCV->m_pMapMatrix	= NULL;
	pOpenCV->iWidth         = 0;
	pOpenCV->iHeight        = 0;
}
 
void OpenCVwarpFilterPlugin::filterClose(void* piParams)
{
	OpenCVStruct* pOpenCV = (OpenCVStruct*)piParams;
	if (pOpenCV->m_pImg != NULL)
	{
		cvReleaseImage(&pOpenCV->m_pImg);
		cvReleaseImage(&pOpenCV->m_pOutputImg);
		cvReleaseMat(&pOpenCV->m_pMapMatrix);
	}
}
 
void OpenCVwarpFilterPlugin::setOperation(unsigned char *destintation,unsigned char *source,
		              long width,long height,long dest_pitch,long src_pitch,long Bpp,void* piParams)
{
	OpenCVStruct* pOpenCV = (OpenCVStruct*)piParams;
	if (! pOpenCV)
		return;
 
 
 
	if ((pOpenCV->m_pImg == NULL) || (pOpenCV->iWidth != width) || (pOpenCV->iHeight != height))
	{
		if (pOpenCV->m_pImg != NULL)
		{
			cvReleaseImage(&pOpenCV->m_pImg);
			cvReleaseImage(&pOpenCV->m_pOutputImg);
			cvReleaseMat(&pOpenCV->m_pMapMatrix);
		}
        pOpenCV->m_pImg = cvCreateImage(cvSize(width,height), IPL_DEPTH_8U, 3);
        pOpenCV->m_pOutputImg = cvCreateImage(cvGetSize(pOpenCV->m_pImg), 8, 3 );
 
		pOpenCV->src_quad[0].x = 150;  pOpenCV->src_quad[0].y = 150;
		pOpenCV->src_quad[1].x = 0;  pOpenCV->src_quad[1].y = (float)height;
		pOpenCV->src_quad[2].x = (float)width;  pOpenCV->src_quad[2].y = 0;
		pOpenCV->src_quad[3].x = (float)width;  pOpenCV->src_quad[3].y = (float)height;
 
		pOpenCV->dst_quad[0].x = 0;  pOpenCV->dst_quad[0].y = 0;
		pOpenCV->dst_quad[1].x = 0;  pOpenCV->dst_quad[1].y = (float)height;
		pOpenCV->dst_quad[2].x = (float)width;  pOpenCV->dst_quad[2].y = 0;
		pOpenCV->dst_quad[3].x = (float)width;  pOpenCV->dst_quad[3].y = (float)height;
 
		pOpenCV->m_pMapMatrix = cvCreateMat(3,3,CV_32FC1);
	}
	pOpenCV->iWidth  = width;
	pOpenCV->iHeight = height;
 
	cvWarpPerspectiveQMatrix(pOpenCV->src_quad, pOpenCV->dst_quad, pOpenCV->m_pMapMatrix);
 
	memcpy(pOpenCV->m_pImg->imageData,source,pOpenCV->m_pImg->imageSize);
    cvWarpPerspective(pOpenCV->m_pImg, pOpenCV->m_pOutputImg, pOpenCV->m_pMapMatrix, CV_WARP_FILL_OUTLIERS,  cvScalarAll(1));
	memcpy(destintation,pOpenCV->m_pOutputImg->imageData,pOpenCV->m_pOutputImg->imageSize);
}
 
Q_EXPORT_PLUGIN2(FilterInterface, OpenCVwarpFilterPlugin)
 
// Header
#ifndef EXTRAFILTERSPLUGIN_H
#define EXTRAFILTERSPLUGIN_H
 
#include <QObject>
#include <QString>
#include "interfaceImageFilter.h"
 
class OpenCVwarpFilterPlugin : public QObject, public FilterInterface
{
    Q_OBJECT
    Q_INTERFACES(FilterInterface)
 
public:
    QString filterName() const;
	unsigned long filterParamsSize() const;
	void filterInit(void* piParams);
	void filterClose(void* piParams);
    void setOperation(unsigned char *dest,unsigned char *src,
		                      long width,long height,long dest_pitch,long src_pitch,long Bpp,void* piParams);
 
};
 
#endif

// code
#include <QtGui>
#include <math.h>
#include <stdlib.h>
#include "OpenCVwarpFilter.h"
//
#include "cv.h"
#include "cxcore.h"

#pragma message("automatic link to OpenCV libs")
#pragma comment(lib,"cv210.lib")
#pragma comment(lib,"cxcore210.lib")

struct OpenCVStruct
{
int             iWidth;
int             iHeight;
IplImage        *m_pImg;        // OpenCV Images
IplImage        *m_pOutputImg;
CvMat           *m_pMapMatrix;  // OpenCV Matrix
CvPoint2D32f     src_quad[4];
CvPoint2D32f     dst_quad[4];
};

QString OpenCVwarpFilterPlugin::filterName() const
{
    return "piWarp_OpenCV";
}


unsigned long OpenCVwarpFilterPlugin::filterParamsSize() const
{
    return sizeof(OpenCVStruct);
}

void OpenCVwarpFilterPlugin::filterInit(void* piParams)
{
	OpenCVStruct* pOpenCV = (OpenCVStruct*)piParams;
	pOpenCV->m_pImg		    = NULL;
	pOpenCV->m_pOutputImg	= NULL;
	pOpenCV->m_pMapMatrix	= NULL;
	pOpenCV->iWidth         = 0;
	pOpenCV->iHeight        = 0;
}

void OpenCVwarpFilterPlugin::filterClose(void* piParams)
{
	OpenCVStruct* pOpenCV = (OpenCVStruct*)piParams;
	if (pOpenCV->m_pImg != NULL)
	{
		cvReleaseImage(&pOpenCV->m_pImg);
		cvReleaseImage(&pOpenCV->m_pOutputImg);
		cvReleaseMat(&pOpenCV->m_pMapMatrix);
	}
}

void OpenCVwarpFilterPlugin::setOperation(unsigned char *destintation,unsigned char *source,
		              long width,long height,long dest_pitch,long src_pitch,long Bpp,void* piParams)
{
	OpenCVStruct* pOpenCV = (OpenCVStruct*)piParams;
	if (! pOpenCV)
		return;

	

	if ((pOpenCV->m_pImg == NULL) || (pOpenCV->iWidth != width) || (pOpenCV->iHeight != height))
	{
		if (pOpenCV->m_pImg != NULL)
		{
			cvReleaseImage(&pOpenCV->m_pImg);
			cvReleaseImage(&pOpenCV->m_pOutputImg);
			cvReleaseMat(&pOpenCV->m_pMapMatrix);
		}
        pOpenCV->m_pImg = cvCreateImage(cvSize(width,height), IPL_DEPTH_8U, 3);
        pOpenCV->m_pOutputImg = cvCreateImage(cvGetSize(pOpenCV->m_pImg), 8, 3 );

		pOpenCV->src_quad[0].x = 150;  pOpenCV->src_quad[0].y = 150;
		pOpenCV->src_quad[1].x = 0;  pOpenCV->src_quad[1].y = (float)height;
		pOpenCV->src_quad[2].x = (float)width;  pOpenCV->src_quad[2].y = 0;
		pOpenCV->src_quad[3].x = (float)width;  pOpenCV->src_quad[3].y = (float)height;

		pOpenCV->dst_quad[0].x = 0;  pOpenCV->dst_quad[0].y = 0;
		pOpenCV->dst_quad[1].x = 0;  pOpenCV->dst_quad[1].y = (float)height;
		pOpenCV->dst_quad[2].x = (float)width;  pOpenCV->dst_quad[2].y = 0;
		pOpenCV->dst_quad[3].x = (float)width;  pOpenCV->dst_quad[3].y = (float)height;

		pOpenCV->m_pMapMatrix = cvCreateMat(3,3,CV_32FC1);
	}
	pOpenCV->iWidth  = width;
	pOpenCV->iHeight = height;

	cvWarpPerspectiveQMatrix(pOpenCV->src_quad, pOpenCV->dst_quad, pOpenCV->m_pMapMatrix);

	memcpy(pOpenCV->m_pImg->imageData,source,pOpenCV->m_pImg->imageSize);
    cvWarpPerspective(pOpenCV->m_pImg, pOpenCV->m_pOutputImg, pOpenCV->m_pMapMatrix, CV_WARP_FILL_OUTLIERS,  cvScalarAll(1));
	memcpy(destintation,pOpenCV->m_pOutputImg->imageData,pOpenCV->m_pOutputImg->imageSize);
}

Q_EXPORT_PLUGIN2(FilterInterface, OpenCVwarpFilterPlugin)

// Header
#ifndef EXTRAFILTERSPLUGIN_H
#define EXTRAFILTERSPLUGIN_H

#include <QObject>
#include <QString>
#include "interfaceImageFilter.h"

class OpenCVwarpFilterPlugin : public QObject, public FilterInterface
{
    Q_OBJECT
    Q_INTERFACES(FilterInterface)

public:
    QString filterName() const;
	unsigned long filterParamsSize() const;
	void filterInit(void* piParams);
	void filterClose(void* piParams);
    void setOperation(unsigned char *dest,unsigned char *src,
		                      long width,long height,long dest_pitch,long src_pitch,long Bpp,void* piParams);

};

#endif

To copy to clipboard, switch view to plain text mode 

[pavanbarot]

Hi josiahbryan,

Thanks For Sharing an Awesome Example. i have little problem with this application can u please help me.
i got following output:

640 is false
320 is true

Listing capture properties...

CV_CAP_PROP_FRAME_WIDTH 0
CV_CAP_PROP_FRAME_HEIGHT 0

CV_CAP_PROP_FPS 0
CV_CAP_PROP_FOURCC 4.29497e+09

CV_CAP_PROP_BRIGHTNESS 0
CV_CAP_PROP_CONTRAST 0

CV_CAP_PROP_SATURATION 0
CV_CAP_PROP_HUE 0

Done

Settings 320x240: 0
Attempting to set frame rate...

Error: 0
Starting to track

About to start the capture thread

Started the capture thread

E: Imagebuffer received a null image

E: Imagebuffer received a null image

E: Imagebuffer received a null image

Hi josiahbryan,

Thanks For Sharing an Awesome Example. i have little problem with this application can u please help me.
i got following output:

640 is false
320 is true

Listing capture properties...

CV_CAP_PROP_FRAME_WIDTH 0
CV_CAP_PROP_FRAME_HEIGHT 0

CV_CAP_PROP_FPS 0
CV_CAP_PROP_FOURCC 4.29497e+09

CV_CAP_PROP_BRIGHTNESS 0
CV_CAP_PROP_CONTRAST 0

CV_CAP_PROP_SATURATION 0
CV_CAP_PROP_HUE 0

Done

Settings 320x240: 0
Attempting to set frame rate...

Error: 0
Starting to track

About to start the capture thread

Started the capture thread

E: Imagebuffer received a null image

E: Imagebuffer received a null image

E: Imagebuffer received a null image