Selective Guidance Normal Filter for Geometric Texture Removal.

Area of Science:

• Computer Graphics
• Geometric Modeling
• Image Processing

Background:

• Conventional 3D mesh filters face a trade-off between texture removal and structure preservation.
• Existing methods often result in artifacts like remnant textures and distorted shapes due to fixed patch-based filtering.

Purpose of the Study:

• To develop a novel mesh filtering technique that simultaneously removes geometric textures and preserves geometric structures of 3D surfaces.
• To introduce a specialized filter that overcomes the limitations of conventional approaches by avoiding the texture-vs-structure trade-off.

Main Methods:

• Proposed a selective guidance normal filter (SGNF) adapting the Relative Total Variation (RTV) to a maximal/minimal scheme (mmRTV).
• The mmRTV quantifies patch flatness, enabling the selection of adaptive patches aligned with facets.
• Utilized adaptive patches to derive selective guidance normals for filtering, distinguishing between texture removal (maximal RTV) and structure preservation (minimal RTV).

Main Results:

• The SGNF effectively smooths geometric textures while preserving essential geometric structures.
• Experimental results demonstrate visual and numerical performance comparable to state-of-the-art mesh filters.
• The mmRTV approach shows applicability beyond mesh filtering, including bas-relief modeling and image texture removal.

Conclusions:

• The SGNF offers a specialized solution for 3D surface processing, effectively balancing texture removal and structure preservation.
• The proposed mmRTV scheme provides a robust and versatile method for enhancing geometric data and images.
• This approach advances 3D surface filtering by providing a non-trade-off solution for detailed appearance and intrinsic properties.