Firewire camera input for Haskell

This repository builds on the bindings-dc1394 library, a low level Haskell bindings for libdc1394. It provides the ability to parse captured firewire images into arrays in the accelerate library, to be processed in parallel on multicore CPUs and GPUs. It also provides the ability to parse captured firewire images into OpenCV image structures in the CV OpenCV Haskell binding. In both cases, the getFrame function should be used in conjunction with withFirewireCamera from the bindings-dc1394 fork, see Section Development.

Project structure

Image processing on CPUs and GPUs with Accelerate

The accelerate mapping is in Data.Array.Accelerate.IO.Firewire and images are captured with:

getFrame :: Camera -> IO (Either Error (Array DIM3 Word8))

Image processing on CPUs with OpenCV

The OpenCV mapping is in CV.Image.IO and images are captured with:

getFrame :: Camera -> IO (Maybe (Image RGB D8))

Example

Full examples of both are in examples/. Here’s one:

main :: IO ()
main = do
    -- map a firewire frame into an accelerate array
    withFirewireCamera ISO_400
                       Mode_640x480_RGB8 Rate_30
                       4
                       defaultFlags $ \camera -> do
      Right frame <- A.getFrame camera
      let rgbaImg = (A.map A.packRGBA32
                    . reverseThen3dTo2d 480 640
                    . use) frame
          computation = (A.map A.rgba32OfLuminance .fst . canny 0.7 0.9) rgbaImg
          -- run Canny edge detection on a GPU
          newImg  = CUDA.run computation
      A.writeImageToBMP "out_accelerate.bmp" newImg  

    -- map a firewire frame into an OpenCV image
    withFirewireCamera ISO_400
                       Mode_640x480_RGB8 Rate_30
                       4
                       defaultFlags $ \camera -> do
      -- just save OpenCV image to file
      fromJust <$> CV.getFrame camera >>= CV.saveImage "out_openCV.png"

Development