background-remover

背景移除器

使用多种检测方法移除图像背景。

功能

• 基于颜色移除：移除纯色背景
• 边缘检测：检测主体边缘并进行移除
• GrabCut 算法：交互式提取前景
• 批量处理：处理多张图像
• 透明输出：导出带有 alpha 通道的图像
• 背景替换：替换为颜色或图像

快速开始

from background_remover import BackgroundRemover

remover = BackgroundRemover()

# Simple removal
remover.load("photo.jpg")
remover.remove_background()
remover.save("photo_transparent.png")

# Remove specific color
remover.load("product.jpg")
remover.remove_color((255, 255, 255), tolerance=30)  # Remove white
remover.save("product_clean.png")

# Replace background
remover.load("portrait.jpg")
remover.remove_background()
remover.replace_background(color=(0, 120, 255))  # Blue background
remover.save("portrait_blue.png")

CLI 用法

# Remove background (auto-detect)
python background_remover.py --input photo.jpg --output result.png

# Remove specific color
python background_remover.py --input image.jpg --color "255,255,255" --tolerance 30 -o clean.png

# Use GrabCut method
python background_remover.py --input photo.jpg --method grabcut -o result.png

# Replace background with color
python background_remover.py --input photo.jpg --replace-color "0,120,255" -o result.png

# Replace background with image
python background_remover.py --input photo.jpg --replace-image bg.jpg -o result.png

# Batch process
python background_remover.py --batch input_folder/ --output-dir output/ --method edge

API 参考

BackgroundRemover 类

class BackgroundRemover:
    def __init__(self)

    # Loading
    def load(self, filepath: str) -> 'BackgroundRemover'
    def load_array(self, array: np.ndarray) -> 'BackgroundRemover'

    # Removal Methods
    def remove_background(self, method: str = "auto") -> 'BackgroundRemover'
    def remove_color(self, color: Tuple, tolerance: int = 20) -> 'BackgroundRemover'
    def remove_edges(self, threshold: int = 50) -> 'BackgroundRemover'
    def grabcut(self, rect: Tuple = None, iterations: int = 5) -> 'BackgroundRemover'

    # Background Operations
    def replace_background(self, color: Tuple = None, image: str = None) -> 'BackgroundRemover'
    def add_shadow(self, offset: Tuple = (5, 5), blur: int = 10) -> 'BackgroundRemover'

    # Refinement
    def refine_edges(self, feather: int = 2) -> 'BackgroundRemover'
    def expand_mask(self, pixels: int = 2) -> 'BackgroundRemover'
    def contract_mask(self, pixels: int = 2) -> 'BackgroundRemover'

    # Output
    def save(self, filepath: str, quality: int = 95) -> str
    def get_image(self) -> Image
    def get_mask(self) -> Image

    # Batch Processing
    def batch_process(self, input_dir: str, output_dir: str,
                     method: str = "auto") -> List[str]

移除方法

自动检测

# Automatically choose best method
remover.remove_background(method="auto")

基于颜色移除

# Remove white background
remover.remove_color((255, 255, 255), tolerance=30)

# Remove green screen
remover.remove_color((0, 255, 0), tolerance=50)

# Remove any solid color
remover.remove_color((200, 200, 200), tolerance=40)

边缘检测

# Use edge detection to find subject
remover.remove_edges(threshold=50)

GrabCut（OpenCV）

# Full image GrabCut
remover.grabcut(iterations=5)

# With bounding rectangle hint
remover.grabcut(rect=(50, 50, 400, 300), iterations=10)

背景替换

纯色

remover.remove_background()
remover.replace_background(color=(255, 255, 255))  # White
remover.replace_background(color=(0, 0, 0))        # Black
remover.replace_background(color=(135, 206, 235))  # Sky blue

图像背景

remover.remove_background()
remover.replace_background(image="office_bg.jpg")

透明（默认）

remover.remove_background()
remover.save("transparent.png")  # PNG preserves alpha

边缘优化

# Soften edges with feathering
remover.refine_edges(feather=3)

# Expand mask to include more area
remover.expand_mask(pixels=2)

# Contract mask for tighter crop
remover.contract_mask(pixels=2)

工作流示例

产品摄影

remover = BackgroundRemover()

# Remove white studio background
remover.load("product_photo.jpg")
remover.remove_color((255, 255, 255), tolerance=25)
remover.refine_edges(feather=2)
remover.save("product_transparent.png")

人像编辑

remover = BackgroundRemover()

# Remove background from portrait
remover.load("portrait.jpg")
remover.grabcut(iterations=8)
remover.refine_edges(feather=3)

# Add professional background
remover.replace_background(color=(220, 220, 220))
remover.add_shadow(offset=(5, 5), blur=15)
remover.save("portrait_professional.jpg")

绿幕移除

remover = BackgroundRemover()

remover.load("greenscreen_video_frame.jpg")
remover.remove_color((0, 255, 0), tolerance=60)
remover.replace_background(image="virtual_bg.jpg")
remover.save("composited.jpg")

批量处理

remover = BackgroundRemover()

processed = remover.batch_process(
    input_dir="product_photos/",
    output_dir="processed/",
    method="color",
    color=(255, 255, 255),
    tolerance=30
)

print(f"Processed {len(processed)} images")

输出格式

• PNG：保留透明度（推荐）
• WEBP：文件更小，支持 alpha
• JPEG：不支持透明度（请搭配 replace_background 使用）

获得最佳效果的技巧

1. 白色／纯色背景：使用 remove_color() 方法
2. 复杂背景：使用 grabcut() 方法
3. 高对比度主体：边缘检测效果很好
4. 人像：使用 GrabCut 并优化边缘
5. 产品照片：使用颜色移除并进行羽化

局限性

• 主体与背景对比度较高时效果最佳
• 复杂的头发／毛发边缘可能需要手动修饰
• 透明或半透明主体较难处理
• 非常杂乱的背景可能需要手动协助

依赖项

• pillow>=10.0.0
• opencv-python>=4.8.0
• numpy>=1.24.0
• scikit-image>=0.21.0