imshow ROI has different size than ROI.shape

Question

Evening all!

My questions are the following:

1.The shape of roi is only supposed to be 200,200. Here it is showing 313,313 and Im not sure why.

2.When I switch frame for roi as the argument to imshow(), the result is also not the correct size of the roi. What I get is this output:

from imutils.video import VideoStream
import argparse
import cv2
import numpy as np
import imutils
import time

ap = argparse.ArgumentParser(description="Pass the .mp4 file and optional buffer size as arguments.")
ap.add_argument("-video", type=str, help="Path to the video file") .
ap.add_argument("-buffer", type=int, default=64, help="Optional max buffer size")
args = vars(ap.parse_args())
video = cv2.VideoCapture(args["video"])
time.sleep(2.0)
offsetx = 38
offsety = 5 

while (video.isOpened()): 
    ret, frame = video.read()
    frame = imutils.resize(frame, width=1280)
    (screenheight, screenwidth) = frame.shape[:2]
    mid_y = screenheight/2
    mid_x = screenwidth /2
    roi_x = (mid_x + offsetx) - 100
    roi_y = (mid_y + offsety) - 100
    roi = frame[int(roi_y):int(roi_x),int(roi_y)+200:int(roi_x)+200]
    cv2.rectangle(frame,(int(roi_x),int(roi_y)),(int(roi_x)+200,int(roi_y)+200),(0,255,0),3)
    print(int(roi_x),int(roi_y),int(roi_x)+200,int(roi_y)+200)
    print(roi.shape)

    cv2.imshow('frame',roi)
        if cv2.waitKey(1) & 0xFF == ord('q'):
        break

video.release()
cv2.destroyAllWindows()

The output of this file is:

(313, 313, 3) 578 265 778 465 ...

your roi specification is wrong it should be [y1:y2, x1:x2] not [y1:x1, y2:x2] — fmw42, Sep 17 '19 at 01:28

fmw42 · Accepted Answer · 2019-09-17T02:18:29.220

1

Your roi specification is incorrect. It should be [y1:y2, x1:x2] not [y1:x1, y2:x2]

Here I test on a static image:

import cv2

img = cv2.imread('lena.jpg')
height = img.shape[0]
width = img.shape[1]

mid_y = height/2
mid_x = width/2

offset_y = -50
offset_x = -25
y1 = int(mid_y + offset_y) 
x1 = int(mid_x + offset_x)
rows = 100
cols = 50
y2 = y1+rows
x2 = x1+cols
print(y1, y2, x1, x2)

roi = img[y1:y2, x1:x2]
cv2.rectangle(img, (x1,y1), (x2,y2), (0,255,0), 3)
print(roi.shape)

cv2.imshow("ROI", roi)
cv2.waitKey(0)
cv2.destroyAllWindows()

cv2.imwrite("lena_roi.jpg", roi)

Prints:

78 178 103 153
(100, 50, 3)

Result:

edited Sep 17 '19 at 02:18

answered Sep 17 '19 at 01:43

fmw42

46,825
10
62
80

Thank you, this works. I kept looking for the syntax you mentioned on this site but could not find it https://opencv-python-tutroals.readthedocs.io/en/latest/py_tutorials/py_core/py_basic_ops/py_basic_ops.html#basic-ops. Would be nice if it was written out syntax-wise as above. – noc_coder Sep 17 '19 at 01:55
See https://docs.opencv.org/3.4/d1/d9d/classcv_1_1MatOp.html#a4a39d30a977f213f196420244fe1708b and https://docs.opencv.org/3.3.0/d3/df2/tutorial_py_basic_ops.html – fmw42 Sep 17 '19 at 02:24
Hello, as a followup question, I'm not entirely sure how this roi function is related to the frame. Are you referencing it for the ranges only? – noc_coder Sep 18 '19 at 03:28
It is being used to crop the raster image. In your case the file would be loaded from your video frame. In my case, I simply read in an image file from disk. The ranges are the ranges of coordinates used to crop the image that is loaded from your video frame. – fmw42 Sep 18 '19 at 04:23

imshow ROI has different size than ROI.shape

1 Answers1