import numpy as np
import cv2

def check_color(image):
    """
    Determine whether there are colors in non black, gray, white, and other gray areas in an RGB color image。
    params：
    image -- np.array
    return：
    True -- Colors with non black, gray, white, and other grayscale areas
    False -- Images are all grayscale areas
    """
    gray_image = np.dot(image[...,:3], [0.299, 0.587, 0.114])
    n=0
    for i in range(image.shape[0]):
        for j in range(image.shape[1]):
            color = image[i, j]
            color_distance = np.sum((color - gray_image[i, j])**2)
            if color_distance > 100:
                n+=1
    # gt 10
    if n>10:
        return True
    return False

def is_ignore(region_img, ignore_bubble = 0):
    """
    Principle: Normally, white bubbles and their text boxes are mostly white, while black bubbles and their text boxes are mostly black. We calculate the ratio of white or black pixels around the text block to the total pixels, and judge whether the area is a normal bubble area or not. Based on the value of the --ignore-bubble parameter, if the ratio is greater than the base value and less than (100-base value), then it is considered a non-bubble area.
    The normal range for ignore-bubble is 1-50, and other values are considered not input. The recommended value for ignore-bubble is 10. The smaller it is, the more likely it is to recognize normal bubbles as image text and skip them. The larger it is, the more likely it is to recognize image text as normal bubbles.

    Assuming ignore-bubble = 10
    The text block is surrounded by white if it is <10, and the text block is very likely to be a normal white bubble.
    The text block is surrounded by black if it is >90, and the text block is very likely to be a normal black bubble.
    Between 10 and 90, if there are black and white spots around it, the text block is very likely not a normal bubble, but an image.

    The input parameter is the image data of the text block processed by OCR.
    Calculate the ratio of black or white pixels in the four rectangular areas formed by taking 2 pixels from the edges of the four sides of the image.
    Return the overall ratio. If it is between ignore_bubble and (100-ignore_bubble), skip it.

    last determine if there is color, consider the colored text as invalid information and skip it without translation
    """
    if ignore_bubble<1 or ignore_bubble>50:
        return  False
    _, binary_raw_mask = cv2.threshold(region_img, 127, 255, cv2.THRESH_BINARY)
    height, width = binary_raw_mask.shape[:2]

    total=0
    val0=0

    val0+= sum(binary_raw_mask[0:2, 0:width].ravel() == 0)
    total+= binary_raw_mask[0:2, 0:width].size

    val0+= sum(binary_raw_mask[height-2:height, 0:width].ravel() == 0)
    total+= binary_raw_mask[height-2:height, 0:width].size

    val0+= sum(binary_raw_mask[2:height-2, 0:2].ravel() == 0)
    total+= binary_raw_mask[2:height-2, 0:2].size

    val0+= sum(binary_raw_mask[2:height-2, width-2:width].ravel() == 0)
    total += binary_raw_mask[2:height-2, width-2:width].size

    ratio = round( val0 / total, 6)*100
    # ignore
    if ratio>=ignore_bubble and ratio<=(100-ignore_bubble):
        return True
    # To determine if there is color, consider the colored text as invalid information and skip it without translation
    if check_color(region_img):
        return True
    return False