How to crop object in the image on the particular background color in opencv-python?

问题

I want to crop to an object in my image, so that only the colored object remains. How can I do it in python in the most efficient way?

Basically the image has black (0,0,0) background but different colors for an object. I want to crop to the object in order to drop the useless background.

I know cv2 has the resize() function but they cannot detect whether it is the background or not. I can also loop the whole image to find the position but that is too slow.

回答1:

Finally I have find an API to do the work.

use cv2.findContours() to get the position of the object from the mask image (an object with a corresponding color) and directly cut it with numpy.

def cut_object(rgb_image,mask_image,object_color):
    """This function is used to cut a specific object from the pair RGB/mask image."""
    rgb_image=cv2.imread(rgb_image)
    mask_image=cv2.imread(mask_image)
    mask_image=cv2.cvtColor(mask_image,cv2.COLOR_BGR2RGB)

    # Create mask image with the only object
    object_mask_binary=cv2.inRange(mask_image,object_color,object_color)
    object_mask=cv2.bitwise_and(mask_image,mask_image,mask=object_mask_binary)

    # Detect the position of the object
    object_contour=cv2.cvtColor(object_mask,cv2.COLOR_BGR2GRAY)
    object_position,c=cv2.findContours(object_contour,cv2.RETR_TREE,cv2.CHAIN_APPROX_SIMPLE)
    object_position=np.array(object_position).squeeze()
    hmin,hmax=object_position[:,:1].min(),object_position[:,:1].max()
    wmin,wmax=object_position[:,1:2].min(),object_position[:,1:2].max()

    # Cut the object from the RGB image
    crop_rgb=rgb_image[wmin:wmax,hmin:hmax]

    return crop_rgb

来源：https://stackoverflow.com/questions/55495807/how-to-crop-object-in-the-image-on-the-particular-background-color-in-opencv-pyt