LayerMask: MaskEdgeShrink

Documentation

• Class name: LayerMask: MaskEdgeShrink
• Category: 😺dzNodes/LayerMask
• Output node: False

The MaskEdgeShrink node is designed to process image masks by shrinking their edges, optionally inverting the masks, and applying a level of softness to the edge transitions. It allows for detailed control over the edge modification process, including the ability to preserve certain edge details while reducing others, making it suitable for creating stylized or more precise mask boundaries in image editing and compositing tasks.

Input types

Required

• mask
  • The input mask to be processed. It serves as the primary data for edge modification, determining the areas of the image where edge shrinking will be applied.
  • Comfy dtype: MASK
  • Python dtype: torch.Tensor
• invert_mask
  • A boolean flag indicating whether the input mask should be inverted before processing. Inversion can be useful for operations targeting the area outside the original mask.
  • Comfy dtype: BOOLEAN
  • Python dtype: bool
• shrink_level
  • Specifies the intensity of the edge shrinking operation. Higher values result in more significant edge reduction.
  • Comfy dtype: INT
  • Python dtype: int
• soft
  • Determines the softness of the edge transitions after shrinking, allowing for smoother or more abrupt changes at the mask edges.
  • Comfy dtype: INT
  • Python dtype: int
• edge_shrink
  • Controls the amount of edge that is specifically targeted for shrinking, enabling precise manipulation of edge width.
  • Comfy dtype: INT
  • Python dtype: int
• edge_reserve
  • Defines the extent to which original edge details are preserved during the shrinking process, allowing for a balance between edge reduction and detail retention.
  • Comfy dtype: INT
  • Python dtype: int

Optional

Output types

• mask
  • Comfy dtype: MASK
  • The processed mask with shrunken edges, returned as a tensor. This output reflects the modifications applied to the original mask, including any inversion, softening, and detail preservation specified.
  • Python dtype: torch.Tensor

Usage tips

• Infra type: GPU
• Common nodes: unknown

Source code

class MaskEdgeShrink:

    def __init__(self):
        pass

    @classmethod
    def INPUT_TYPES(self):

        return {
            "required": {
                "mask": ("MASK", ),  #
                "invert_mask": ("BOOLEAN", {"default": True}),  # 反转mask
                "shrink_level": ("INT", {"default": 4, "min": 0, "max": 16, "step": 1}),
                "soft": ("INT", {"default": 6, "min": 0, "max": 64, "step": 1}),
                "edge_shrink": ("INT", {"default": 1, "min": 0, "max": 999, "step": 1}),
                "edge_reserve": ("INT", {"default": 25, "min": 0, "max": 100, "step": 1}),  # 透明度
            },
            "optional": {
            }
        }

    RETURN_TYPES = ("MASK",)
    RETURN_NAMES = ("mask",)
    FUNCTION = 'mask_edge_shrink'
    CATEGORY = '😺dzNodes/LayerMask'

    def mask_edge_shrink(self, mask, invert_mask, shrink_level, soft, edge_shrink, edge_reserve):

        l_masks = []
        ret_masks = []

        if mask.dim() == 2:
            mask = torch.unsqueeze(mask, 0)

        for m in mask:
            if invert_mask:
                m = 1 - m
            l_masks.append(tensor2pil(torch.unsqueeze(m, 0)).convert('L'))

        glow_range = shrink_level * soft
        blur = 12

        for i in range(len(l_masks)):
            _mask = l_masks[i]
            _canvas = Image.new('RGB', size=_mask.size, color='black')
            _layer = Image.new('RGB', size=_mask.size, color='white')
            loop_grow = glow_range
            inner_mask = _mask
            for x in range(shrink_level):
                _color = step_color('#FFFFFF', '#000000', shrink_level, x)
                glow_mask = expand_mask(image2mask(inner_mask), -loop_grow, blur / (x+0.1))  #扩张，模糊
                # 合成
                color_image = Image.new("RGB", _layer.size, color=_color)
                alpha = tensor2pil(mask_invert(glow_mask)).convert('L')
                _glow = chop_image(_layer, color_image, 'subtract', int(step_value(1, 100, shrink_level, x)))
                _layer.paste(_glow, mask=alpha)
                loop_grow = loop_grow - int(glow_range / shrink_level)
            # 合成layer
            _edge = tensor2pil(expand_mask(image2mask(_mask), -edge_shrink, 0)).convert('RGB')
            _layer = chop_image(_layer, _edge, 'normal', edge_reserve)
            _layer.paste(_canvas, mask=ImageChops.invert(_mask))

            ret_masks.append(image2mask(_layer))

        log(f"{NODE_NAME} Processed {len(ret_masks)} mask(s).", message_type='finish')
        return (torch.cat(ret_masks, dim=0),)