Enhancing Image Matting in Real-World Scenes with Mask-Guided Iterative Refinement

TL;DR

This paper introduces Mask2Alpha, an iterative framework that improves real-world image matting by leveraging self-supervised vision transformer features, mask-guided feature selection, and sparse convolution-based refinement, achieving state-of-the-art results.

Contribution

The paper presents a novel iterative refinement framework combining semantic priors, mask-guided feature selection, and sparse convolution optimization for enhanced image matting.

Findings

Achieves state-of-the-art performance on real-world datasets.

Effectively recovers high-resolution details through progressive refinement.

Enhances semantic understanding and instance awareness in complex scenes.

Abstract

Real-world image matting is essential for applications in content creation and augmented reality. However, it remains challenging due to the complex nature of scenes and the scarcity of high-quality datasets. To address these limitations, we introduce Mask2Alpha, an iterative refinement framework designed to enhance semantic comprehension, instance awareness, and fine-detail recovery in image matting. Our framework leverages self-supervised Vision Transformer features as semantic priors, strengthening contextual understanding in complex scenarios. To further improve instance differentiation, we implement a mask-guided feature selection module, enabling precise targeting of objects in multi-instance settings. Additionally, a sparse convolution-based optimization scheme allows Mask2Alpha to recover high-resolution details through progressive refinement,from low-resolution semantic passes to high-resolution sparse reconstructions. Benchmarking across various real-world datasets, Mask2Alpha consistently achieves state-of-the-art results, showcasing its effectiveness in accurate and efficient image matting.