Face Tracking with CAMShift using OpenCV/SimpleCV

CAMShift stands for Continuously Adaptive Mean Shift. It is the basis for the face-tracking algorithm in OpenCV. It combines the basic Mean Shift algorithm with an adaptive region-sizing step. The kernel is a simple step function applied to a skin-probability map. The skin probability of each image pixel is based on color using a method called histogram backprojection. Color is represented as Hue from the HSV color model.

Face Tracking with CAMShift

import cv2
import cv2.cv as cv

def camshift_tracking(img1, img2, bb):
        hsv = cv2.cvtColor(img1, cv.CV_BGR2HSV)
        mask = cv2.inRange(hsv, np.array((0., 60., 32.)), np.array((180., 255., 255.)))
        x0, y0, w, h = bb
        x1 = x0 + w -1
        y1 = y0 + h -1
        hsv_roi = hsv[y0:y1, x0:x1]
        mask_roi = mask[y0:y1, x0:x1]
        hist = cv2.calcHist( [hsv_roi], [0], mask_roi, [16], [0, 180] )
        cv2.normalize(hist, hist, 0, 255, cv2.NORM_MINMAX);
        hist_flat = hist.reshape(-1)
        prob = cv2.calcBackProject([hsv,cv2.cvtColor(img2, cv.CV_BGR2HSV)], [0], hist_flat, [0, 180], 1)
        prob &= mask
        term_crit = ( cv2.TERM_CRITERIA_EPS | cv2.TERM_CRITERIA_COUNT, 10, 1 )
        new_ellipse, track_window = cv2.CamShift(prob, bb, term_crit)
        return track_window

def face_track():
    cap = cv2.VideoCapture(0)
    img = cap.read()
    bb = (125,125,200,100) # get bounding box from some method
    while True:
            img1 = cap.read()
            bb = camshift(img1, img, bb)
            img = img1
            #draw bounding box on img1
        except KeyboardInterrupt:

In SimpleCV,

from SimpleCV import Image, Camera
cam = Camera()
img = cam.getImage()
# Move your face in some direction
img1 = cam.getImage()
bb = (100,200,100,100) # get Bounding Box from some method
new_bb = img1.track(img,"camshift",bb)
# do this in loop.

It seems that CAMShift is quite efficient and fast. I’ll add Lucas Kanede Tracker, Median Flow Tracker, Forward-Backward Error Tracker soon. And to top it all, I will add OpenTLD to SimpleCV.

P.S. Down with throat infection. Waiting for SimpleCV t-shirt and my Mid-Term evaluation.

About these ads

5 thoughts on “Face Tracking with CAMShift using OpenCV/SimpleCV

    • Yeah, it’s the really simple to learn and implement. It’s really easy to test some new algorithm of Computer Vision. And as it is Python, syntax is relatively easier than C++ and no memory management, garbage collection issues. And since it uses OpenCV python bindings, it has speed and is faster than MATLAB and some other Python based libraries.

  1. hai jay,i m dinesh how to run the open cv code…what u have given,,it produces some error…becoz i m haviing one header file which is having all headers & libraries…whether it is right or wrong.

Leave a Reply

Please log in using one of these methods to post your comment:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s