1hew-ComfyUI-1hewNodes/web/docs/1hew_DetectRemoveBGRefine/en.md

Detect Remove BG Refine - RMBG Mask Post-Refinement

Node Purpose: Detect Remove BG Refine is designed as a post-processing node for RMBG-1.4-style outputs. It takes the original image plus model mask, performs alpha refinement, anti-aliasing, optional vector hard-edge shaping, and decontamination, then outputs optimized RGBA and refined mask.

Inputs

Name	Port	Type	Default	Range	Description
image	required	IMAGE	-	-	Original source image (must be the original, not model-processed image).
mask	required	MASK	-	-	Alpha mask produced by RMBG model output.
type	-	COMBO	bitmap	bitmap / vector	Edge mode: soft bitmap edges or harder vector-like contour edges.
subject_protect	-	FLOAT	0.85	0.0~1.0	Foreground core protection strength.
feather	-	FLOAT	1.0	0.0~64.0	Feather control (mapped internally to smoothing intensity).
decolor_edge	-	FLOAT	1.0	0.0~1.0	Edge decontamination strength.

Outputs

Name	Type	Description
image	IMAGE	Standard RGBA output (alpha as the 4th channel, no premultiply).
mask	MASK	Refined alpha mask.

Features

• Alpha refinement first: bilateral/Gaussian smoothing, foreground-core protection, and edge anti-aliasing.
• Optional vector hardening: vector mode reconstructs harder contour shapes after refinement.
• Fixed background estimation policy: white/black priority with auto fallback.
• Decontamination always uses original RGB with refined alpha.

Typical Usage

• RMBG polishing: feed mask from Detect Remove BG, and pass the original image to this node.
• Hard-edge assets (logos/graphics): use type=vector and increase subject_protect if needed.
• Natural portrait edges: use type=bitmap with moderate feather and decolor_edge.

Notes & Tips

• Do not feed model-processed images back into this node; it can amplify edge pollution and color shifts.
• Internal policy is fixed by design; there are no extra bg_mode, min_area_ratio, or premultiply parameters.