Image Matting

Module: Computer Vision | Difficulty: Advanced

Matting Equation

where is foreground, is background, .

Loss Function

Trimap-Based Matting

GrabCut

Iterative graph cut for foreground extraction.

import torch
import torch.nn as nn

class MattingNet(nn.Module):
    def __init__(self):
        super().__init__()
        self.encoder = nn.Sequential(
            nn.Conv2d(4, 64, 3, padding=1), nn.ReLU(True),
            nn.Conv2d(64, 128, 3, 2, 1), nn.ReLU(True),
            nn.Conv2d(128, 256, 3, 2, 1), nn.ReLU(True),
        )
        self.decoder = nn.Sequential(
            nn.ConvTranspose2d(256, 128, 4, 2, 1), nn.ReLU(True),
            nn.ConvTranspose2d(128, 64, 4, 2, 1), nn.ReLU(True),
            nn.Conv2d(64, 1, 3, padding=1), nn.Sigmoid(),
        )
    
    def forward(self, x, trimap):
        inp = torch.cat([x, trimap], dim=1)
        feat = self.encoder(inp)
        return self.decoder(feat)

Key Takeaways

• Matting estimates per-pixel alpha values
• Trimap guides the matting algorithm to the transition region
• Deep matting methods can work without explicit trimaps