rui40000andClaude Opus 4.8
5d3834503e
【修复】角色被格线切断（脚、手杖、飘起的斗篷被削掉） 新增 expand_beyond_cell（默认开启）：格子只用来判定「这是哪个方向」， 角色的实际范围由它自身的连通区域决定，按质心归属确保邻居不混入。 实测 8/8 方向的裁剪框边缘 alpha 从 1.00（内容顶到边界=被切断） 降到 0.00，S 方向高度 326→356、E 方向宽度 150→188 把缺的部分找了回来。 代价是需要两遍扫描（先求全序列并集框再提取），耗时 4.7s→14.9s。 【规则】每个参数都必须有中文 tooltip，作为以后的统一约定 全仓库 26 个节点 169 个参数，此前缺 115 个，现已 100% 覆盖。 tooltip 写「怎么调」而不只是「是什么」：给取值区间的实际影响、 推荐值与踩坑提示（如 OpenAI/ZenMux 的地址栏不能带 :// ， 素材拆分节点用于动画序列时顺序会漂移等）。 Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
128 lines
5.0 KiB
Python
128 lines
5.0 KiB
Python
import numpy as np
|
|
import torch
|
|
|
|
def to_numpy_uint8(img_t): # [H,W,C] float(0..1) -> np.uint8
|
|
img = img_t.detach().cpu().numpy()
|
|
img = (np.clip(img, 0.0, 1.0) * 255.0).round().astype(np.uint8)
|
|
return img
|
|
|
|
def to_tensor_image(img_np): # np.uint8 HWC -> torch [H,W,C] float(0..1)
|
|
t = torch.from_numpy(img_np.astype(np.float32) / 255.0)
|
|
return t
|
|
|
|
def histogram_match(source, template):
|
|
# 逐通道直方图匹配
|
|
src = source.copy()
|
|
for c in range(src.shape[2]):
|
|
s = src[..., c].ravel()
|
|
t = template[..., c].ravel()
|
|
s_values, bin_idx, s_counts = np.unique(s, return_inverse=True, return_counts=True)
|
|
t_values, t_counts = np.unique(t, return_counts=True)
|
|
s_quantiles = np.cumsum(s_counts).astype(np.float64) / s.size
|
|
t_quantiles = np.cumsum(t_counts).astype(np.float64) / t.size
|
|
interp_t_values = np.interp(s_quantiles, t_quantiles, t_values)
|
|
src[..., c] = interp_t_values[bin_idx].reshape(src.shape[:2])
|
|
return src
|
|
|
|
class ColorMatcherNode:
|
|
"""颜色匹配节点：将移动图像的颜色分布匹配到参考图像"""
|
|
@classmethod
|
|
def INPUT_TYPES(cls):
|
|
return {
|
|
"required": {
|
|
"reference_image": ("IMAGE", {
|
|
"tooltip": "参考图：提供目标色调，本身不会被改动。"
|
|
}),
|
|
"moving_image": ("IMAGE", {
|
|
"tooltip": "待校色图：它的颜色分布会被改造成参考图的样子。"
|
|
}),
|
|
"match_method": (["histogram", "mean_std", "none"], {
|
|
"default": "histogram",
|
|
"tooltip": "匹配算法：\n"
|
|
"histogram = 逐通道直方图匹配，色调贴合最紧，\n"
|
|
" 但两图内容差异大时容易出偏色\n"
|
|
"mean_std = 只对齐均值与标准差，效果温和、不易翻车\n"
|
|
"none = 不做匹配，仅用于对照原图"
|
|
}),
|
|
"blend_factor": ("FLOAT", {
|
|
"default": 1.0, "min": 0.0, "max": 1.0, "step": 0.01,
|
|
"tooltip": "匹配强度：结果在原图与完全匹配之间线性混合。\n"
|
|
"0 = 保持原样，1 = 完全套用参考图色调。\n"
|
|
"匹配过头显得失真时降到 0.5~0.8 往往更自然。"
|
|
}),
|
|
}
|
|
}
|
|
|
|
RETURN_TYPES = ("IMAGE", "STRING")
|
|
RETURN_NAMES = ("颜色匹配后图像", "匹配信息")
|
|
FUNCTION = "match_colors"
|
|
CATEGORY = "Rui-Node🐶/图像调节🎨"
|
|
|
|
def match_colors(self, reference_image, moving_image, match_method, blend_factor):
|
|
B = reference_image.shape[0]
|
|
assert moving_image.shape[0] == B, "batch 大小必须一致"
|
|
|
|
out_images = []
|
|
match_info = []
|
|
|
|
for i in range(B):
|
|
ref_img = reference_image[i] # [H,W,C]
|
|
mov_img = moving_image[i]
|
|
|
|
if match_method == "none":
|
|
match_info.append(f"Batch {i}: 跳过颜色匹配")
|
|
out_images.append(mov_img.unsqueeze(0))
|
|
continue
|
|
|
|
ref_np = to_numpy_uint8(ref_img)
|
|
mov_np = to_numpy_uint8(mov_img)
|
|
|
|
if match_method == "histogram":
|
|
# 直方图匹配
|
|
matched_np = histogram_match(mov_np, ref_np)
|
|
method_name = "直方图匹配"
|
|
elif match_method == "mean_std":
|
|
# 均值标准差匹配
|
|
matched_np = self.mean_std_match(mov_np, ref_np)
|
|
method_name = "均值标准差匹配"
|
|
else:
|
|
matched_np = mov_np
|
|
method_name = "无匹配"
|
|
|
|
# 混合原图和匹配后的图像
|
|
if blend_factor < 1.0:
|
|
matched_np = (matched_np * blend_factor + mov_np * (1 - blend_factor)).astype(np.uint8)
|
|
|
|
match_info.append(f"Batch {i}: {method_name}，混合系数: {blend_factor:.2f}")
|
|
matched_tensor = to_tensor_image(matched_np)
|
|
out_images.append(matched_tensor.unsqueeze(0))
|
|
|
|
matched_images = torch.cat(out_images, dim=0)
|
|
info_str = " | ".join(match_info)
|
|
|
|
return (matched_images, info_str)
|
|
|
|
def mean_std_match(self, source, template):
|
|
"""均值标准差颜色匹配"""
|
|
src = source.astype(np.float32)
|
|
tpl = template.astype(np.float32)
|
|
|
|
matched = src.copy()
|
|
for c in range(src.shape[2]):
|
|
src_mean = np.mean(src[..., c])
|
|
src_std = np.std(src[..., c])
|
|
tpl_mean = np.mean(tpl[..., c])
|
|
tpl_std = np.std(tpl[..., c])
|
|
|
|
if src_std > 0:
|
|
matched[..., c] = (src[..., c] - src_mean) * (tpl_std / src_std) + tpl_mean
|
|
|
|
return np.clip(matched, 0, 255).astype(np.uint8)
|
|
|
|
NODE_CLASS_MAPPINGS = {
|
|
"ColorMatcher": ColorMatcherNode
|
|
}
|
|
|
|
NODE_DISPLAY_NAME_MAPPINGS = {
|
|
"ColorMatcher": "颜色匹配器 / Color Matcher"
|
|
} |