smaa

SMAA, short for Subpixel Morphological Anti-Aliasing, is a real-time post-processing technique used in computer graphics to reduce jagged edges on rendered images. It aims to produce high-quality edge anti-aliasing while preserving fine textures and avoiding excessive blur, offering an appealing balance between image quality and performance.

The SMAA pipeline generally comprises three stages. First, edge detection identifies potential screen-space edges with subpixel precision by analyzing color and luminance differences across neighboring pixels. This step seeks to locate true geometry edges while suppressing edges created by texture detail or lighting. Second, pattern classification or pattern search examines the local neighborhood around detected edges to determine the edge type and how best to blend colors along it, using a library of morphological patterns to distinguish true edges from textured color variations. Third, neighborhood blending reconstructs the final image by blending pixel colors along the detected edges with carefully computed weights, producing smoother transitions without overly blurring the rest of the image.

SMAA is a post-processing technique, meaning it does not modify scene geometry or require multi-sampling of geometry. It typically offers higher-quality edge anti-aliasing than some faster methods like FXAA while demanding less computational power than multisampling approaches such as MSAA. Variants of SMAA incorporate temporal information to further reduce temporal aliasing and shimmering, at the cost of increased complexity. Since its introduction, SMAA has been implemented in various graphics engines and has influenced subsequent anti-aliasing developments in real-time rendering.