确保ROI裁剪坐标(x，y，w，h)是有效的，并且在Python OpenCV的范围内

Question

给定从图像中裁剪ROI的(x,y,w,h)坐标，如何确保给定的坐标有效？例如：

image = cv2.imread('1.jpeg')
x,y,w,h = 0,0,300,300
ROI = image[y:y+h,x:x+w]
cv2.imshow('ROI', ROI)
cv2.waitKey()

如果(x,y,w,h)坐标无效，它将引发以下错误：

错误: C:\projects\opencv-python\opencv\modules\highgui\src\window.cpp:350:错误：(-215:断言失败)函数'cv::imshow‘中的size.width>0 && size.height>0

我试图在裁剪ROI之前编写一个函数来验证坐标。目前，我的一些检查是为了确保：

1. (x,y,w,h)要么是int，要么是float，type
2. x和y是>=，0
3. w和h是> 0

。

有时它仍然会抛出错误，我遗漏了哪些检查？

示例图像：

代码：

import cv2

def validate_ROI_coordinates(coordinates):
    # (x,y) is top left coordinates
    # Top right corner is is (x + w)
    # Bottom left corner is (y + h) 
    
    x,y,w,h = coordinates
    
    # Ensure its a number, not boolean or string type
    def int_or_float(s):
        try:
            i = int(s)
            return True
        except ValueError:
            try:
                f = float(s)
                return True
            except:
                return False
    
    def within_bounds(x,y,w,h):
        # Ensure that x and y are >= 0
        if x >= 0 and y >= 0:
            # Ensure w and h are > 0 ( can be any positive number)
            if w > 0 and h > 0:
                return True
        else:
            return False
    
    if all(int_or_float(value) for value in coordinates) and within_bounds(x,y,w,h):
        return True
    else:
        return False

image = cv2.imread('1.jpeg')
print(image.shape)
x,y,w,h = 500,0,6600,300
coordinates = (x,y,w,h)

if validate_ROI_coordinates(coordinates):
    ROI = image[y:y+h,x:x+w]
    cv2.imshow('ROI', ROI)
    cv2.waitKey()
else:
    print('Invalid ROI coordinates')

Answer 1

可以使用图像分辨率检查坐标是否在图像范围内：

# get resolution and coordinates
height, width = image.shape[:-1]
xmin, ymin, w, h = coordinates
xmax = xmin + w
ymax = ymin + h

# fetch roi
if (xmin  >= 0) and (ymin >= 0) and (xmax < width) and (ymax < height):
  roi = image[ymin:ymax, xmin:xmax]