去水印python

在 Python 中去除水印可以根據水印類型（圖片、文字、視頻）和復雜度選擇不同的工具和算法。以下是常見場景的解決方案：

一、靜態圖片去水印

1. 基礎圖像處理（簡單水印）

```python

from PIL import Image, ImageEnhance

def remove_watermark(input_path, output_path):

img = Image.open(input_path)

檢測水印區域（需預先知道位置）

watermark = img.crop((left, top, right, bottom))

img.paste(watermark, (0,0), watermark)

img.save(output_path)

```

2. 智能去水印（深度學習）

使用預訓練模型（需安裝 pytorch）：

```python

import torch

from remove.bg import Remove

def remove_bg(input_path):

remove = Remove()

with open(input_path, "rb") as f:

img = remove.remove(f)

img.save("output.png")

```

3. OCR + 圖像修復（文字水印）

```python

import easyocr

from PIL import ImageEnhance

def remove_text_watermark(input_path):

OCR 識別文字區域

reader = easyocr.Reader(['ch_sim', 'en'])

result = reader.readtext(input_path)

 獲取水印坐標并修復
img = Image.open(input_path)
for det in result:
if det[4] > 0.7:  置信度判斷
left, top, right, bottom, conf = det[:5]
img = img.crop((left, top, right, bottom))
img.save("output.png")

```

二、視頻去水印

1. FFmpeg + Python

```python

import subprocess

def remove_video_watermark(input_path, output_path):

cmd = [

'ffmpeg', '-i', input_path,

'-vf', 'movie=watermark.png@0.5', 添加半透明水印（示例）

output_path

]

subprocess.run(cmd)

```

2. 實時去水印（需GPU）

```python

import torch

from videoai import VideoAI

def real_time_remove(input_path):

videoai = VideoAI('watermark_remove_model')

videoai.remove_watermark(input_path, output_path)

```

三、API 服務集成

1. Remove.bg API

   ```python

   import requests

def remove_bg_api(input_path):

with open(input_path, "rb") as f:

response = requests.post(

"https://api.remove.bg/v1.0/removebg",

files={"image": f