如何在python中从具有多个遮罩的单个图像创建每个图像包含一个实例遮罩的单独图像

Question

I have the following image:

Each mask is for an individual instance. I want to have separate images which will contain only one instance mask per image. Masks are not overlapping strictly. The outputs will be as like following:

Thank You.

Answer 1

I have an answer to this. The following code segment creates the desired output:

def create_separate_mask(path):
    # get all the masks
    for mask_file in glob.glob(path + '/*mask.png'): 
        mask = cv2.imread(mask_file, 1)
        # get masks labelled with different values
        label_im, nb_labels = ndimage.label(mask) 

        for i in range(nb_labels):

            # create an array which size is same as the mask but filled with 
            # values that we get from the label_im. 
            # If there are three masks, then the pixels are labeled 
            # as 1, 2 and 3.

            mask_compare = np.full(np.shape(label_im), i+1) 

            # check equality test and have the value 1 on the location of each mask
            separate_mask = np.equal(label_im, mask_compare).astype(int) 

            # replace 1 with 255 for visualization as rgb image

            separate_mask[separate_mask == 1] = 255 
            base=os.path.basename(mask_file)

            # give new name to the masks

            file_name = os.path.splitext(base)[0]
            file_copy = os.path.join(path, file_name + "_" + str(i+1) +".png") 
            cv2.imwrite(file_copy, separate_mask)