getCalibratedRGBPixels not correct??

[ofTheo]

See http://theo.tw/i/__calibratedpixels-issue.png Looks like the math is doing the opposite.

[kylemcdonald]

did this just break in a recent revision, or have you noticed this before?

[ofTheo]

this is with the latest version.
I hadn't checked it sense the new ofxKinectCall files were added.

Here is some sample code for reproducing the issue - move your mouse to compare the two images.

//--------------------------------------------------------------
void testApp::setup()
{
    //kinect.init(true);  //shows infrared image
    kinect.init();
    kinect.setVerbose(true);
    kinect.open();

    colorImg.allocate(kinect.width, kinect.height);
    grayImage.allocate(kinect.width, kinect.height);
    grayThresh.allocate(kinect.width, kinect.height);
    grayThreshFar.allocate(kinect.width, kinect.height);

    nearThreshold = 50;
    farThreshold  = 180;
    bThreshWithOpenCV = true;

    ofSetFrameRate(60);

    drawPC = false;
}

//--------------------------------------------------------------
void testApp::update()
{
    ofBackground(100, 100, 100);
    kinect.update();

    grayImage.setFromPixels(kinect.getDepthPixels(), kinect.width, kinect.height);
    colorImg.setFromPixels(kinect.getCalibratedRGBPixels(), kinect.width, kinect.height);
}

//--------------------------------------------------------------
void testApp::draw(){

    ofEnableAlphaBlending();

    ofSetColor(255, 255, 255, 255); 
    colorImg.draw(0, 0);

    ofSetColor(255, 255, 255, ((float)mouseX/ofGetWidth()) * 255.0);    
    grayImage.draw(0,0);

    ofSetColor(255, 255, 255);
    ofDrawBitmapString("accel is: " + ofToString(kinect.getMksAccel().x, 2) + " / " 
                                    + ofToString(kinect.getMksAccel().y, 2) + " / "
                                    + ofToString(kinect.getMksAccel().z, 2), 20, 658 );

    char reportStr[1024];
    sprintf(reportStr, "using opencv threshold = %i (press spacebar)\nset near threshold %i (press: + -)\nset far threshold %i (press: < >) num blobs found %i, fps: %f",bThreshWithOpenCV, nearThreshold, farThreshold, contourFinder.nBlobs, ofGetFrameRate());
    ofDrawBitmapString(reportStr, 20, 690);
    ofDrawBitmapString("tilt angle: " + ofToString(angle),20,670);
    ofDrawBitmapString("press p to switch between images and point cloud",20,680);
}

[kylemcdonald]

yeah i definitely see this.

http://oi53.tinypic.com/i43g5t.jpg

arturo is the one who added the calibrated color stuff in the first place i think so he's probably going to be able to solve this more easily.

i know he's been working a lot with calibrating the depth and color images, but i thought this part was already solved... so i'm not sure what's going on.

[ofTheo]

cool thanks for the sanity check! :)

[ofTheo]

just talked to arturo ( who is not able to connect to github ) it seems this function is meant for the particle demo and not for correctly mapping the depth image to the color image which makes sense now that I think about it.

might still need testing for the particle demo as I think Kyle's current code renders the particles in b&w.

[danomatika]

What's the status on this?

[danomatika]

It seems the calibration numbers vary slightly form Kinect to Kinect so the numbers we have now are not exactly perfect. We would need to provide a method to generate a calibration file using a pattern like openCV.

See here for an example: http://nicolas.burrus.name/index.php/Research/KinectRgbDemoV4?from=Research.KinectRgbDemoV2#tocLink10

[danomatika]

Check the develop branch to try the new libfreenect which includes calibration ... it's working!

[ofTheo]

awesome - when can we merge the current develop into master? btw the multi kinect stuff works great dan - thanks for hooking that up.