Extracting raw data from Kinect

edit

You might want to convert the sensor_msgs/Image datatype to an OpenCV image. The cv_bridge tutorial for Python shows how to do so: http://www.ros.org/wiki/cv_bridge/Tutorials/ConvertingBetweenROSImagesAndOpenCVImagesPython

The below code is an example from our SVN ( http://code.google.com/p/albany-ros-pkg ) that captures an image, blurs it, and displays it:

#!/usr/bin/env python
""" 
  Example code of how to convert ROS images to OpenCV's cv::Mat
  This is the solution to HW2, using Python.

  See also cv_bridge tutorials: 
    http://www.ros.org/wiki/cv_bridge
"""

import roslib; roslib.load_manifest('ex_vision')
import rospy

import cv
from cv_bridge import CvBridge, CvBridgeError

from std_msgs.msg import String
from sensor_msgs.msg import Image

class image_blur:

  def __init__(self):
    # initialize a node called hw2
    rospy.init_node("hw2")

    # create a window to display results in
    cv.NamedWindow("image_view", 1)

    # part 2.1 of hw2 -- subscribe to a topic called image
    self.image_sub = rospy.Subscriber("image", Image, self.callback)

  def callback(self,data):
    """ This is a callback which recieves images and processes them. """
    # convert image into openCV format
    bridge = CvBridge()
    try:
      # bgr8 is the pixel encoding -- 8 bits per color, organized as blue/green/red
      cv_image = bridge.imgmsg_to_cv(data, "bgr8")
    except CvBridgeError, e:
      # all print statements should use a rospy.log_ form, don't print!
      rospy.loginfo("Conversion failed")

    # we could do anything we want with the image here
    # for now, we'll blur using a median blur
    cv.Smooth(cv_image, cv_image, smoothtype=cv.CV_MEDIAN, param1=31, param2=0, param3=0, param4=0)

    # show the image
    cv.ShowImage("image_view", cv_image)
    cv.WaitKey(3)


if __name__ == '__main__':
    image_blur()
    try:  
        rospy.spin()
    except KeyboardInterrupt:
        print "Shutting down"
    cv.DestroyAllWindows()