Exploring the Interactive Guidance for Unified and Effective Image Matting

TL;DR

This paper introduces UIM, a unified interactive image matting method that effectively handles multiple objects and transparent materials by leveraging diverse user interactions and a two-stage process, achieving state-of-the-art results.

Contribution

The paper proposes a novel unified framework for image matting that combines multiple interaction types and separates transparency prediction from foreground segmentation.

Findings

UIM outperforms existing methods on Composition-1K and synthetic datasets.

The multi-scale attentive fusion module improves boundary accuracy.

Decoupling transparency from foreground segmentation enhances performance on transparent objects.

Abstract

Recent image matting studies are developing towards proposing trimap-free or interactive methods for complete complex image matting tasks. Although avoiding the extensive labors of trimap annotation, existing methods still suffer from two limitations: (1) For the single image with multiple objects, it is essential to provide extra interaction information to help determining the matting target; (2) For transparent objects, the accurate regression of alpha matte from RGB image is much more difficult compared with the opaque ones. In this work, we propose a Unified Interactive image Matting method, named UIM, which solves the limitations and achieves satisfying matting results for any scenario. Specifically, UIM leverages multiple types of user interaction to avoid the ambiguity of multiple matting targets, and we compare the pros and cons of different annotation types in detail. To unify the matting performance for transparent and opaque objects, we decouple image matting into two stages, i.e., foreground segmentation and transparency prediction. Moreover, we design a multi-scale attentive fusion module to alleviate the vagueness in the boundary region. Experimental results demonstrate that UIM achieves state-of-the-art performance on the Composition-1K test set and a synthetic unified dataset.