一、安裝

• 安裝tesseract：brew install tesseract
• 安裝python包：pip3 install pytesseract
• 下載中文模型：下載 https://github.com/tesseract-ocr/tessdata/blob/main/chi_sim.traineddata
  ，到/usr/local/Cellar/tesseract/5.3.3/share/tessdata

二、使用

from PIL import Image
import pytesseract

img = cv2.imread('aaa.png')
text = pytesseract.image_to_string(img, lang="chi_sim")
print(text)

去除噪點

from PIL import Image
import pytesseract
import cv2

img = cv2.imread('aaa.png')
grayimg = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY) # # 轉(zhuǎn)為灰度圖

_, bwimg = cv2.threshold(grayimg, 127, 255, cv2.THRESH_BINARY)  # 二值化

contours, _ = cv2.findContours(bwing, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)  # 尋找輪廓


# 方法一
# 選出面積較小區(qū)域
noise = []
for contour in contours:
    area = cv2.contourArea(contour)
    if area <= 40:
        noise.append(contour)

cv2.fillPoly(bwimg, noise, 0) # 刪除點填充為 黑色

# 方法 二
kernel = np.ones((3, 11), np.uint8)
dilation = cv2.dilate(bwimg, kernel, iterations=1)  # 膨脹

# 膨脹后連在一起
contours, _ = cv2.findContours(dilation, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)  # 尋找輪廓
text_area = max(contours, key=lambda x: cv2.contourArea(x))

# 最小外接矩陣
rect = cv2.minAreaRect(text_area)
box = cv2.boxPoints(rect)
box = np.int0(box)

# 矩陣外填充為黑色
stencil = np.zeros(bwimg.shape).astype(bwimg.dtype)
color = [255, 255, 255]
cv2.fillPoly(stencil, [box], color)
result = cv2.bitwise_and(bwimg, stencil)