Mesh saliency via spectral processing

Mesh saliency is a measure that attempts to capture the importance of a point or local region on a 3D surface mesh in a similar way to human visual perception. The human perceptual system is able to detect visual saliency extraordinarily quickly and reliably, even for novel scenes. The term “saliency” is often considered in the context of bottom-up computations. However, development of bottom-up computational models which simulate this basic intelligent behaviour remains a profound challenge in computer vision and graphics.

Mesh saliency detection methods usually merge perceptual criteria inspired by low-level human visual cues with mathematical measures based on discrete differential geometry, such as curvatures. Overall, however, saliency must efficiently and effectively reflect perceptually important regions on a 3Dmesh, which curvature alone may not capture. While mesh saliency may not outperform mesh curvature as a surface analysis tool in all applications, it provides an alternative approach when processing 3D meshes, based on perceptual mechanisms rather than purely local geometric measures of shape.

The project proposes a novel method for detecting mesh saliency, a perceptuallybased measure of the importance of a local region on a 3D surface mesh. Our method incorporates global considerations by making use of spectral attributes of the mesh, unlike most existing methods which are typically based on local geometric cues. We first consider the properties of the log- Laplacian spectrum of the mesh. Those frequencies which show differences from expected behaviour capture saliency in the frequency domain. Information about these frequencies is considered in the spatial domain at multiple spatial scales to localise the salient features and give the final salient areas.

The effectiveness and robustness of the approach is demonstrated by comparisons to previous approaches on a range of test models. The benefits of the proposed method are further evaluated in applications such as mesh simplification, mesh segmentation, and scan integration, where we show how incorporating mesh saliency can provide improved results.

Publication
Mesh saliency via spectral processing, ACM Transactions on Graphics, Volume 33Issue 1Article No.: 6pp 1–17https://doi.org/10.1145/2530691