Well I could not do it with Wand. I found another library Pillow.
I have a java script code that capture video frame from canvas and convert the webp imge from based64 to binary image and send it using web socket to a server on the server I construct the image and convert it from webp to jpeg and then use OpenCV to process the jpeg image. Here is a sample
code
from PIL import Image
import StringIO
import numpy as np
import cv2
#webpimg is binary webp image received from the websocket
newImg = Image.open(StringIO.StringIO(webpimg)).convert("")
temp = StringIO.StringIO()
newImg.save(temp, "JPEG")
contents = temp.getvalue()
temp.close()
array = np.fromstring(contents, dtype=np.uint8)
jpegimg = cv2.imdecode(array, cv2.CV_LOAD_IMAGE_COLOR)
cv2.imwrite("imgCV.jpeg", img1)