Python实现噪声平面图线条降噪与实体化的技术求助

问题背景 #

对噪声严重的平面图进行线条降噪及实体化处理，已尝试masking、bluring、HoughLinesP及前两者的组合方法。使用HoughLinesP可实现线条实体化，但文字区域线条严重重叠，且调整minLineLength、maxLineGap参数无法解决该问题。

原始处理代码 #

import cv2
import numpy as np
from tkinter import Tk     # from tkinter import Tk for Python 3.x
from tkinter.filedialog import askopenfilename
import os

Tk().withdraw() # 隐藏TK主窗口
filename = askopenfilename() # 打开文件选择对话框
print(filename)

filename3, file_extension = os.path.splitext(filename)

# 读取灰度图
img = cv2.imread(filename, cv2.IMREAD_GRAYSCALE)

# 初始化彩色输出图
out = cv2.cvtColor(img, cv2.COLOR_GRAY2BGR)

# 中值模糊降噪并反转图像
img = 255 - cv2.medianBlur(img, 3)

# 检测并绘制线条
lines = cv2.HoughLinesP(img, 1, np.pi/180, 10, minLineLength=40, maxLineGap=30)
for line in lines:
    for x1, y1, x2, y2 in line:
        cv2.line(out, (x1, y1), (x2, y2), (0, 0, 255), 2)

cv2.imshow('out', out)
cv2.imwrite(filename3+' '+'69'+'.png', out)
cv2.waitKey(0)
cv2.destroyAllWindows()

问题分析 #

文字区域的线条重叠本质是文字笔画被HoughLinesP误识别为线条，由于文字笔画短且密集，单纯调整长度或间隙参数会陷入“过滤文字则丢失有效短线条”的矛盾，需从预处理阶段分离文字与线条，或采用更精准的线条检测逻辑。

解决方案 #

方案1：形态学操作过滤文字区域 #

利用文字与线条的形态差异（线条为长条形，文字为小区域密集笔画），通过长条形核的开运算过滤文字，保留目标线条：

import cv2
import numpy as np
from tkinter import Tk
from tkinter.filedialog import askopenfilename
import os

Tk().withdraw()
filename = askopenfilename()
print(filename)

filename3, file_extension = os.path.splitext(filename)

# 读取灰度图并初始化输出
img = cv2.imread(filename, cv2.IMREAD_GRAYSCALE)
out = cv2.cvtColor(img, cv2.COLOR_GRAY2BGR)

# 反转图像（背景黑，线条白）
img_inv = 255 - img

# 用水平长核开运算，过滤文字保留水平线条
horizontal_kernel = cv2.getStructuringElement(cv2.MORPH_RECT, (30, 1))
img_horizontal = cv2.morphologyEx(img_inv, cv2.MORPH_OPEN, horizontal_kernel, iterations=2)

# 用垂直长核开运算，保留垂直线条
vertical_kernel = cv2.getStructuringElement(cv2.MORPH_RECT, (1, 30))
img_vertical = cv2.morphologyEx(img_inv, cv2.MORPH_OPEN, vertical_kernel, iterations=2)

# 合并水平与垂直线条
img_lines = cv2.addWeighted(img_horizontal, 0.5, img_vertical, 0.5, 0)

# 降噪后检测线条
img_blur = cv2.medianBlur(img_lines, 3)
lines = cv2.HoughLinesP(img_blur, 1, np.pi/180, threshold=20, minLineLength=40, maxLineGap=30)

if lines is not None:
    for line in lines:
        x1, y1, x2, y2 = line[0]
        cv2.line(out, (x1, y1), (x2, y2), (0, 0, 255), 2)

cv2.imshow('Processed Lines', out)
cv2.imwrite(filename3+'_processed.png', out)
cv2.waitKey(0)
cv2.destroyAllWindows()

方案2：轮廓筛选替代HoughLinesP #

通过Canny边缘检测提取轮廓，筛选长条形轮廓（线条），过滤小面积轮廓（文字）：

import cv2
import numpy as np
from tkinter import Tk
from tkinter.filedialog import askopenfilename
import os

Tk().withdraw()
filename = askopenfilename()
print(filename)

filename3, file_extension = os.path.splitext(filename)

img = cv2.imread(filename, cv2.IMREAD_GRAYSCALE)
out = cv2.cvtColor(img, cv2.COLOR_GRAY2BGR)

# 降噪+Canny边缘检测
img_blur = cv2.GaussianBlur(img, (5,5), 0)
edges = cv2.Canny(img_blur, 50, 150)

# 提取轮廓并筛选长条形线条
contours, _ = cv2.findContours(edges, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)
for cnt in contours:
    x, y, w, h = cv2.boundingRect(cnt)
    # 计算长宽比，保留长线条（可根据图像调整阈值）
    aspect_ratio = max(w, h) / min(w, h) if min(w, h) !=0 else 0
    if aspect_ratio > 10 and (w > 30 or h > 30):
        cv2.rectangle(out, (x,y), (x+w, y+h), (0,0,255), 2)

cv2.imshow('Contour-based Lines', out)
cv2.imwrite(filename3+'_contour.png', out)
cv2.waitKey(0)
cv2.destroyAllWindows()

方案3：文字检测屏蔽区域 #

用OCR工具检测文字区域，在HoughLinesP检测时屏蔽这些区域（需安装pytesseract及Tesseract引擎）：

import cv2
import numpy as np
import pytesseract
from tkinter import Tk
from tkinter.filedialog import askopenfilename
import os

# 配置Tesseract路径（根据实际安装路径调整）
pytesseract.pytesseract.tesseract_cmd = r'C:\Program Files\Tesseract-OCR\tesseract.exe'

Tk().withdraw()
filename = askopenfilename()
print(filename)

filename3, file_extension = os.path.splitext(filename)

img = cv2.imread(filename, cv2.IMREAD_GRAYSCALE)
out = cv2.cvtColor(img, cv2.COLOR_GRAY2BGR)

# 检测高置信度文字区域，生成掩码
d = pytesseract.image_to_data(img, output_type=pytesseract.Output.DICT)
mask = np.ones_like(img, dtype=np.uint8) * 255
n_boxes = len(d['text'])

for i in range(n_boxes):
    if int(d['conf'][i]) > 60:
        x, y, w, h = d['left'][i], d['top'][i], d['width'][i], d['height'][i]
        cv2.rectangle(mask, (x, y), (x+w, y+h), 0, -1)

# 应用掩码去除文字区域
img_no_text = cv2.bitwise_and(img, mask)

# 后续线条检测流程
img_inv = 255 - cv2.medianBlur(img_no_text, 3)
lines = cv2.HoughLinesP(img_inv, 1, np.pi/180, 10, minLineLength=40, maxLineGap=30)

if lines is not None:
    for line in lines:
        x1, y1, x2, y2 = line[0]
        cv2.line(out, (x1, y1), (x2, y2), (0, 0, 255), 2)

cv2.imshow('No Text Lines', out)
cv2.imwrite(filename3+'_no_text.png', out)
cv2.waitKey(0)
cv2.destroyAllWindows()

方案选型建议 #

• 形态学操作方案：无额外依赖，适合文字与线条形态差异明显的场景；
• 轮廓筛选方案：自定义性强，可灵活调整轮廓特征过滤无效区域；
• 文字检测方案：适合文字复杂但需精准屏蔽的场景，需额外安装OCR工具。

内容的提问来源于stack exchange，提问作者Luke-McDevitt