H-SPAM: Hierarchical Superpixel Anything Model

2026-04-13

AI Summary

Existing superpixel methods have reached a plateau in segmentation accuracy, often producing irregular shapes and offering only single-scale partitions, which limits their applicability to multiscale vision tasks. This work proposes a unified framework that starts from a fine-grained initial partition and integrates deep features with external object priors through a two-stage region merging strategy. For the first time, it achieves a hierarchical superpixel structure that simultaneously ensures high accuracy, geometric regularity, and full nesting. The hierarchy can be adaptively guided by visual attention or user interaction to preserve salient regions. Experiments demonstrate that the method significantly outperforms existing hierarchical approaches on standard benchmarks, achieving state-of-the-art performance in both accuracy and regularity while matching the best non-hierarchical methods.

Technology Category

• Computer Vision: Segmentation
• Machine Learning: Multi-class/Multi-label Learning & Extreme Classification
• Intelligent Robots: Multimodal Perception & Sensor Fusion

Application Category

• Search and Retrieval-Augmented AI: Efficiency and scalability of Web search engines
• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

Superpixels offer a compact image representation by grouping pixels into coherent regions. Recent methods have reached a plateau in terms of segmentation accuracy by generating noisy superpixel shapes. Moreover, most existing approaches produce a single fixed-scale partition that limits their use in vision pipelines that would benefit multi-scale representations. In this work, we introduce H-SPAM (Hierarchical Superpixel Anything Model), a unified framework for generating accurate, regular, and perfectly nested hierarchical superpixels. Starting from a fine partition, guided by deep features and external object priors, H-SPAM constructs the hierarchy through a two-phase region merging process that first preserves object consistency and then allows controlled inter-object grouping. The hierarchy can also be modulated using visual attention maps or user input to preserve important regions longer in the hierarchy. Experiments on standard benchmarks show that H-SPAM strongly outperforms existing hierarchical methods in both accuracy and regularity, while performing on par with most recent state-of-the-art non-hierarchical methods. Code and pretrained models are available: https://github.com/waldo-j/hspam.

Problem

• superpixels
• hierarchical segmentation
• multi-scale representation
• segmentation accuracy
• image partitioning

Innovation

• hierarchical superpixels
• region merging
• deep features
• object priors
• visual attention