字符提取场景下判断轮廓是否大部分为白色填充的实现方法咨询

我有如下图像：

我可以基于密实度（solidity）、高度、*宽高比（aspect ratio）*提取所有字母，但提取结果同时包含了最左侧的方块。由于该图像由多个字母（及一个方块）水平堆叠组成，我需要判断一个轮廓是否大部分为白色填充。

我使用以下代码从输入图像中提取得到候选字符图像：

from skimage import measure
import numpy as np
import cv2


plate_img = cv2.imread('bla1.png')    
V = cv2.split(cv2.cvtColor(plate_img, cv2.COLOR_BGR2HSV))[2]
thresh = cv2.adaptiveThreshold(V, 255, cv2.ADAPTIVE_THRESH_GAUSSIAN_C, cv2.THRESH_BINARY_INV, 15, 6)
labels = measure.label(thresh, connectivity = 2, background = 0)
charCandidates = np.zeros(thresh.shape, dtype=np.uint8)

characters = []
for label in np.unique(labels):
  # 忽略背景标签
  if label == 0:
    continue
  # 构建标签掩码，仅显示当前标签对应的连通组件，再在掩码中查找轮廓
  labelMask = np.zeros(thresh.shape, dtype=np.uint8)
  labelMask[labels == label] = 255

  cnts, hierarchy = cv2.findContours(labelMask, cv2.RETR_TREE, cv2.CHAIN_APPROX_SIMPLE)

  if len(cnts) > 0:
    # 取掩码中面积最大的轮廓，获取其外接矩形
    c = max(cnts, key=cv2.contourArea)
    (boxX, boxY, boxW, boxH) = cv2.boundingRect(c)

    # 计算组件的宽高比、密实度和高度占比
    aspectRatio = boxW / float(boxH)
    solidity = cv2.contourArea(c) / float(boxW * boxH)
    heightRatio = boxH / float(plate_img.shape[0])

    # 判断轮廓的宽高比、密实度和高度是否符合规则
    keepAspectRatio = aspectRatio < 1.05
    keepSolidity = 0.15 < solidity < 0.875
    keepHeight = 0.5 < heightRatio < 0.95

    # 检查组件是否通过所有校验
    if keepAspectRatio and keepSolidity and keepHeight and boxW >= 6:
      # 计算轮廓的凸包，绘制到候选字符掩码上
      hull = cv2.convexHull(c)
      cv2.drawContours(charCandidates, [hull], -1, 255, -1)

contours = cv2.findContours(charCandidates, cv2.RETR_TREE, cv2.CHAIN_APPROX_SIMPLE)[0]
for c in contours:
  (x, y, w, h) = cv2.boundingRect(c)
  temp = thresh[y:y + h, x:x + w]
  characters.append(temp)
      
return characters

内容的提问来源于stack exchange，提问作者Michael Kročka