Question

or is it entirely dependent on Bumblebee (or any other Point Gray Imaging sensor for that matter)??
I am trying to link the stereo processing API that comes with bumblebee 2 stereo rig to process some offline images. It appears that the triclops stereo function needs some context flag and calibration file in the form input.cal and input.ppm (i.e one image that contains the two stereo rectified images somehow blurry and overlaid). How do I go about obtaining this input.ppm file from other offline images to still be able to use triclops stereo API. And how about the calibration file input.cal, can it be obtained such that it is consistent with the offline images and how ?

Answer 1

Yes, the Triclops API can be used to process offline images. The requirements as you point out are the calibration file (.cal) which comes from the camera and the image file. The calibration file is best extracted at image capture time, though in my experience unless you change the settings it always stays the same. To satisfy the inputs for the Triclops API you’ll also need to use the Point Grey Fly Capture API as it provides the methods for image conversion. My project is doing exactly all of what you are asking. I capture raw pixel interleaved images in the field and then process them later using the FlyCapture and Triclops API to create rectified and disparity images as well as a point cloud from the stereo pair.

Check the Point Grey samples and their API documentation for how to capture the calibration file. Essentially the method is this: flycaptureGetCalibrationFileFromCamera( context, &szCalFile ); Where the context is the ‘camera context’

Your initial image file format will vary the conversion method, but in my case here is how I did the conversion from raw to the stereo using C++. I’m using a bumblebee XB3 color camera with raw images 1280×960 in size and grabbing images from the outer two cameras:

   // Read raw file 
   //pFile = fopen ( "FlyCap.raw" , "rb" );
   pFile = fopen ( argv[1] , "rb" );
   if (pFile==NULL) {fputs ("File error",stderr); exit (1);}

   // Set raw file size 
   ISize = (numCols*numRows*bytesPerPixel);

   // allocate memory to contain the whole file:
   buffer = (unsigned char*) malloc (sizeof(char)*ISize);
   if (buffer == NULL) {fputs ("Memory error",stderr); exit (2);}

   // copy the file into the buffer:
   result = fread (buffer,1,ISize,pFile);
   if (result != ISize) {fputs ("Reading error",stderr); exit (3);}

   // Create the FlyCapture Context for processing
   fe = flycaptureCreateContext( &flycapture );
   _HANDLE_FLYCAPTURE_ERROR( "flycaptureCreateContext()", fe );

   fe = flycaptureSetColorProcessingMethod(flycapture, FLYCAPTURE_HQLINEAR);
   _HANDLE_FLYCAPTURE_ERROR( "flycaptureSetColorProcessingMethod()", fe );

   fe = flycaptureSetColorTileFormat(flycapture, FLYCAPTURE_STIPPLEDFORMAT_GBRG);
   _HANDLE_FLYCAPTURE_ERROR( "flycaptureSetColorTileFormat()", fe );

   //Import the raw image in buffer into FlyCaptureImage structure
   flycaptureImage.iCols = 1280;
   flycaptureImage.iRows = 960;
   flycaptureImage.iNumImages = 2;
   flycaptureImage.bStippled = true;
   flycaptureImage.pixelFormat = FLYCAPTURE_RAW16;   
   flycaptureImage.iRowInc = 2560;
   flycaptureImage.timeStamp.ulSeconds = 100;
   flycaptureImage.timeStamp.ulMicroSeconds = 100;
   flycaptureImage.pData = buffer;

   // Create buffers for holding the color and mono images
   unsigned char* rowIntColor = 
   new unsigned char[ numCols * numRows * flycaptureImage.iNumImages * 4 ];
   unsigned char* rowIntMono = 
   new unsigned char[ numCols * numRows * flycaptureImage.iNumImages ];

   // Create a temporary FlyCaptureImage for preparing the stereo image
   FlyCaptureImage tempColorImage;
   FlyCaptureImage tempMonoImage;

   tempColorImage.pData = rowIntColor;
   tempMonoImage.pData = rowIntMono;


   // Convert the pixel interleaved raw data to row interleaved format
   fe = flycapturePrepareStereoImage( flycapture, flycaptureImage, &tempMonoImage,   &tempColorImage  );
   _HANDLE_FLYCAPTURE_ERROR( "flycapturePrepareStereoImage()", fe );

   //Save side-by-side color stereo image
   fe = flycaptureSaveImage( flycapture, &tempColorImage, stereoimg, FLYCAPTURE_FILEFORMAT_PPM );
   _HANDLE_FLYCAPTURE_ERROR( "flycaptureSaveImage()", fe );  

   printf ("Saving Stereo...\n");

Note that so far I didn’t use the Triclops API at all. The output from the flycaptureSaveImage(); is what feeds into the Triclops methods. There are some options at that point which deal with color vs mono. Take a look at the Point Grey sample that comes with their Triclops API called ‘stereoto3dpoints.cpp’ It takes over from where I left off. I hope this helps you out.