Ask Your Question

Revision history [back]

click to hide/show revision 1
initial version

Live chroma key - optimize script

Hello I've started the following live chroma key in python for a greenscreen:

import numpy as np
import cv2

rows = 800
cols = 600
cap = cv2.VideoCapture(1) #use external cam
ret = cap.set(3,rows)
ret = cap.set(4,cols)

img_back = cv2.imread('background.jpg')
background = img_back[0:cols, 0:rows ]

thresh = 230
blur = 3
final = ""
display=1
kernel = np.ones((5,5),np.uint8)

# create window
cv2.namedWindow("chroma", cv2.WND_PROP_FULLSCREEN)
cv2.setWindowProperty("chroma", cv2.WND_PROP_FULLSCREEN, cv2.cv.CV_WINDOW_FULLSCREEN)

while(True):
    # Capture frame-by-frame
    ret, frame = cap.read()

    frame = cv2.flip(frame,1,frame)

    lab_image = cv2.cvtColor(frame, cv2.COLOR_BGR2LAB)
    l_channel,chan_a,chan_b = cv2.split(lab_image)

    b_channel = cv2.bitwise_not(chan_b)
    a_channel = cv2.add(chan_a,b_channel)
    a_channel = cv2.bitwise_not(a_channel)
    a_channel_inv = cv2.bitwise_not(a_channel)


    # create mask

    im_bw = cv2.threshold(a_channel_inv, thresh, 255, cv2.THRESH_BINARY)[1]
    im_bw = cv2.GaussianBlur(im_bw,(blur,blur),0)


    im_bw = cv2.morphologyEx(im_bw, cv2.MORPH_OPEN, kernel)
    ret ,im_bw = cv2.threshold(im_bw,thresh,255,cv2.THRESH_BINARY)

    im_bw_inv = cv2.bitwise_not(im_bw)
    im_bw_inv = cv2.GaussianBlur(im_bw_inv,(1,1),0)
    im_bw_inv = cv2.morphologyEx(im_bw_inv, cv2.MORPH_CLOSE, kernel)

    roi = cv2.bitwise_and(background,background,mask = im_bw_inv)
    im = cv2.bitwise_and(frame,frame,mask = im_bw)

    # add mask
    dst = cv2.add(roi,im)

    # Display the resulting frame

    k = cv2.waitKey(33)
    if k == 27: #ESC
        break
    if k == ord("+"):
        thresh+=1
    if k == ord("-"):
        thresh-=1

    if k == ord("."):
        blur*=3
    if k == ord(","):
        if blur > 1:
            blur/=3


    if k == ord("1"):
        display = 1
    if k == ord("2"):
        display = 2
    if k == ord("3"):
        display = 3
    if k == ord("4"):
        display = 4
    if k == ord("5"):
        display = 5
    if k == ord("6"):
        display = 6
    if k == ord("7"):
        display = 7
    if k == ord("8"):
        display = 8

    if (display == 1):
        final = dst
    if (display == 2):
        final = im
    if (display == 3):
        final = roi
    if (display == 4):
        final = im_bw
    if (display == 5):
        final = im_bw_inv
    if (display == 6):
        final = chan_a
    if (display == 7):
        final = chan_b
    if (display == 8):
        final = l_channel

    #display image
    cv2.putText(dst,"Thresh (+/-): " + str(thresh), (20,40), cv2.FONT_HERSHEY_SIMPLEX,0.5, 0)
    cv2.putText(dst,"Blur (./,): " + str(blur), (20,60), cv2.FONT_HERSHEY_SIMPLEX,0.5, 0)
    cv2.imshow('chroma', final)

# When everything done, release the capture
cap.release()
cv2.destroyAllWindows()

It's working but since I'm a new to opencv I think it can be optimized and be better (especially in the small areas).

You can use the number keys to switch between the different masks/images and +,.- to change some settings (blur and threshold)

Any thoughts about the script? Looking forward for some optimization!

Best Peter

Live chroma key - optimize script

Hello I've started the following live chroma key in python for a greenscreen:

import numpy as np
import cv2

rows = 800
cols = 600
cap = cv2.VideoCapture(1) #use external cam
ret = cap.set(3,rows)
ret = cap.set(4,cols)

img_back = cv2.imread('background.jpg')
background = img_back[0:cols, 0:rows ]

thresh = 230
blur = 3
final = ""
display=1
kernel = np.ones((5,5),np.uint8)

# create window
cv2.namedWindow("chroma", cv2.WND_PROP_FULLSCREEN)
cv2.setWindowProperty("chroma", cv2.WND_PROP_FULLSCREEN, cv2.cv.CV_WINDOW_FULLSCREEN)

while(True):
    # Capture frame-by-frame
    ret, frame = cap.read()

    frame = cv2.flip(frame,1,frame)

    lab_image = cv2.cvtColor(frame, cv2.COLOR_BGR2LAB)
    l_channel,chan_a,chan_b = cv2.split(lab_image)

    b_channel = cv2.bitwise_not(chan_b)
    a_channel = cv2.add(chan_a,b_channel)
    a_channel = cv2.bitwise_not(a_channel)
    a_channel_inv = cv2.bitwise_not(a_channel)


    # create mask

    im_bw = cv2.threshold(a_channel_inv, thresh, 255, cv2.THRESH_BINARY)[1]
    im_bw = cv2.GaussianBlur(im_bw,(blur,blur),0)


    im_bw = cv2.morphologyEx(im_bw, cv2.MORPH_OPEN, kernel)
    ret ,im_bw = cv2.threshold(im_bw,thresh,255,cv2.THRESH_BINARY)

    im_bw_inv = cv2.bitwise_not(im_bw)
    im_bw_inv = cv2.GaussianBlur(im_bw_inv,(1,1),0)
    im_bw_inv = cv2.morphologyEx(im_bw_inv, cv2.MORPH_CLOSE, kernel)

    roi = cv2.bitwise_and(background,background,mask = im_bw_inv)
    im = cv2.bitwise_and(frame,frame,mask = im_bw)

    # add mask
    dst = cv2.add(roi,im)

    # Display the resulting frame

    k = cv2.waitKey(33)
    if k == 27: #ESC
        break
    if k == ord("+"):
        thresh+=1
    if k == ord("-"):
        thresh-=1

    if k == ord("."):
        blur*=3
    if k == ord(","):
        if blur > 1:
            blur/=3


    if k == ord("1"):
        display = 1
    if k == ord("2"):
        display = 2
    if k == ord("3"):
        display = 3
    if k == ord("4"):
        display = 4
    if k == ord("5"):
        display = 5
    if k == ord("6"):
        display = 6
    if k == ord("7"):
        display = 7
    if k == ord("8"):
        display = 8

    if (display == 1):
        final = dst
    if (display == 2):
        final = im
    if (display == 3):
        final = roi
    if (display == 4):
        final = im_bw
    if (display == 5):
        final = im_bw_inv
    if (display == 6):
        final = chan_a
    if (display == 7):
        final = chan_b
    if (display == 8):
        final = l_channel

    #display image
    cv2.putText(dst,"Thresh (+/-): " + str(thresh), (20,40), cv2.FONT_HERSHEY_SIMPLEX,0.5, 0)
    cv2.putText(dst,"Blur (./,): " + str(blur), (20,60), cv2.FONT_HERSHEY_SIMPLEX,0.5, 0)
    cv2.imshow('chroma', final)

# When everything done, release the capture
cap.release()
cv2.destroyAllWindows()

It's working but since I'm a new to opencv I think it can be optimized and be better (especially in the small areas).

You can use the number keys to switch between the different masks/images and +,.- to change some settings (blur and threshold)

Any thoughts about the script? Looking forward for some optimization!

Best Peter

Live chroma key - optimize script

Hello I've started the following live chroma key in python for a greenscreen:

import numpy as np
import cv2

rows = 800
cols = 600
cap = cv2.VideoCapture(1) #use external cam
ret = cap.set(3,rows)
ret = cap.set(4,cols)

img_back = cv2.imread('background.jpg')
background = img_back[0:cols, 0:rows ]

thresh = 230
blur = 3
final = ""
display=1
kernel = np.ones((5,5),np.uint8)

# create window
cv2.namedWindow("chroma", cv2.WND_PROP_FULLSCREEN)
cv2.setWindowProperty("chroma", cv2.WND_PROP_FULLSCREEN, cv2.cv.CV_WINDOW_FULLSCREEN)

while(True):
    # Capture frame-by-frame
    ret, frame = cap.read()

    frame = cv2.flip(frame,1,frame)

    lab_image = cv2.cvtColor(frame, cv2.COLOR_BGR2LAB)
    l_channel,chan_a,chan_b = cv2.split(lab_image)

    b_channel = cv2.bitwise_not(chan_b)
    a_channel = cv2.add(chan_a,b_channel)
    a_channel = cv2.bitwise_not(a_channel)
    a_channel_inv = cv2.bitwise_not(a_channel)


    # create mask

    im_bw = cv2.threshold(a_channel_inv, thresh, 255, cv2.THRESH_BINARY)[1]
    im_bw = cv2.GaussianBlur(im_bw,(blur,blur),0)


    im_bw = cv2.morphologyEx(im_bw, cv2.MORPH_OPEN, kernel)
    ret ,im_bw = cv2.threshold(im_bw,thresh,255,cv2.THRESH_BINARY)

    im_bw_inv = cv2.bitwise_not(im_bw)
    im_bw_inv = cv2.GaussianBlur(im_bw_inv,(1,1),0)
    im_bw_inv = cv2.morphologyEx(im_bw_inv, cv2.MORPH_CLOSE, kernel)

    roi = cv2.bitwise_and(background,background,mask = im_bw_inv)
    im = cv2.bitwise_and(frame,frame,mask = im_bw)

    # add mask
    dst = cv2.add(roi,im)

    # Display the resulting frame

    k = cv2.waitKey(33)
    if k == 27: #ESC
        break
    if k == ord("+"):
        thresh+=1
    if k == ord("-"):
        thresh-=1

    if k == ord("."):
        blur*=3
    if k == ord(","):
        if blur > 1:
            blur/=3


    if k == ord("1"):
        display = 1
    if k == ord("2"):
        display = 2
    if k == ord("3"):
        display = 3
    if k == ord("4"):
        display = 4
    if k == ord("5"):
        display = 5
    if k == ord("6"):
        display = 6
    if k == ord("7"):
        display = 7
    if k == ord("8"):
        display = 8

    if (display == 1):
        final = dst
    if (display == 2):
        final = im
    if (display == 3):
        final = roi
    if (display == 4):
        final = im_bw
    if (display == 5):
        final = im_bw_inv
    if (display == 6):
        final = chan_a
    if (display == 7):
        final = chan_b
    if (display == 8):
        final = l_channel

    #display image
    cv2.putText(dst,"Thresh (+/-): " + str(thresh), (20,40), cv2.FONT_HERSHEY_SIMPLEX,0.5, 0)
    cv2.putText(dst,"Blur (./,): " + str(blur), (20,60), cv2.FONT_HERSHEY_SIMPLEX,0.5, 0)
    cv2.imshow('chroma', final)

# When everything done, release the capture
cap.release()
cv2.destroyAllWindows()

It's working but since I'm a new to opencv I think it can be optimized and be better (especially in the small areas).

You can use the number keys to switch between the different masks/images and +,.- to change some settings (blur and threshold)

Any thoughts about the script? Looking forward for some optimization!

Best Peter

Live chroma key - optimize script

Hello I've started the following live chroma key in python for a greenscreen:

import numpy as np
import cv2

rows = 800
cols = 600
cap = cv2.VideoCapture(1) #use external cam
ret = cap.set(3,rows)
ret = cap.set(4,cols)

img_back = cv2.imread('background.jpg')
background = img_back[0:cols, 0:rows ]

thresh = 230
blur = 3
final = ""
display=1
kernel = np.ones((5,5),np.uint8)

# create window
cv2.namedWindow("chroma", cv2.WND_PROP_FULLSCREEN)
cv2.setWindowProperty("chroma", cv2.WND_PROP_FULLSCREEN, cv2.cv.CV_WINDOW_FULLSCREEN)

while(True):
    # Capture frame-by-frame
    ret, frame = cap.read()

    frame = cv2.flip(frame,1,frame)

    lab_image = cv2.cvtColor(frame, cv2.COLOR_BGR2LAB)
    l_channel,chan_a,chan_b = cv2.split(lab_image)

    b_channel = cv2.bitwise_not(chan_b)
    a_channel = cv2.add(chan_a,b_channel)
    a_channel = cv2.bitwise_not(a_channel)
    a_channel_inv = cv2.bitwise_not(a_channel)


    # create mask

    im_bw = cv2.threshold(a_channel_inv, thresh, 255, cv2.THRESH_BINARY)[1]
    im_bw = cv2.GaussianBlur(im_bw,(blur,blur),0)


    im_bw = cv2.morphologyEx(im_bw, cv2.MORPH_OPEN, kernel)
    ret ,im_bw = cv2.threshold(im_bw,thresh,255,cv2.THRESH_BINARY)

    im_bw_inv = cv2.bitwise_not(im_bw)
    im_bw_inv = cv2.GaussianBlur(im_bw_inv,(1,1),0)
    im_bw_inv = cv2.morphologyEx(im_bw_inv, cv2.MORPH_CLOSE, kernel)

    roi = cv2.bitwise_and(background,background,mask = im_bw_inv)
    im = cv2.bitwise_and(frame,frame,mask = im_bw)

    # add mask
    dst = cv2.add(roi,im)

    # Display the resulting frame

    k = cv2.waitKey(33)
    if k == 27: #ESC
        break
    if k == ord("+"):
        thresh+=1
    if k == ord("-"):
        thresh-=1

    if k == ord("."):
        blur*=3
    if k == ord(","):
        if blur > 1:
            blur/=3


    if k == ord("1"):
        display = 1
    if k == ord("2"):
        display = 2
    if k == ord("3"):
        display = 3
    if k == ord("4"):
        display = 4
    if k == ord("5"):
        display = 5
    if k == ord("6"):
        display = 6
    if k == ord("7"):
        display = 7
    if k == ord("8"):
        display = 8

    if (display == 1):
        final = dst
    if (display == 2):
        final = im
    if (display == 3):
        final = roi
    if (display == 4):
        final = im_bw
    if (display == 5):
        final = im_bw_inv
    if (display == 6):
        final = chan_a
    if (display == 7):
        final = chan_b
    if (display == 8):
        final = l_channel

    #display image
    cv2.putText(dst,"Thresh (+/-): " + str(thresh), (20,40), cv2.FONT_HERSHEY_SIMPLEX,0.5, 0)
    cv2.putText(dst,"Blur (./,): " + str(blur), (20,60), cv2.FONT_HERSHEY_SIMPLEX,0.5, 0)
    cv2.imshow('chroma', final)

# When everything done, release the capture
cap.release()
cv2.destroyAllWindows()

It's working but since I'm a new to opencv I think it can be optimized and be better (especially in the small areas).

You can use the number keys to switch between the different masks/images and +,.- to change some settings (blur and threshold)

Any thoughts about the script? Looking forward for some optimization!

Best Peter