51 lines
1.7 KiB
Markdown
51 lines
1.7 KiB
Markdown
# Documentation
|
|
- Class name: AutoContrast
|
|
- Category: image/postprocessing
|
|
- Output node: False
|
|
- Repo Ref: https://github.com/Jordach/comfy-plasma.git
|
|
|
|
该节点通过拉伸图像的强度值范围来增强图像的对比度,使其跨越一个期望的范围,提高了图像的视觉吸引力和清晰度,同时不会改变其基本特征。
|
|
|
|
# Input types
|
|
## Required
|
|
- IMAGE
|
|
- 输入图像,其对比度将由节点进行调整,作为增强过程的基础。
|
|
- Comfy dtype: IMAGE
|
|
- Python dtype: PIL.Image
|
|
## Optional
|
|
- cutoff
|
|
- 该参数控制应用于图像的对比度增强程度,较高的值会导致对比度变化更加明显。
|
|
- Comfy dtype: FLOAT
|
|
- Python dtype: float
|
|
- min_value
|
|
- 该参数设置图像中不受对比度调整影响的最小值,保留了阴影部分的细节。
|
|
- Comfy dtype: INT
|
|
- Python dtype: int
|
|
|
|
# Output types
|
|
- IMAGE
|
|
- 输出是对比度提高的图像,适合进一步处理或显示。
|
|
- Comfy dtype: IMAGE
|
|
- Python dtype: PIL.Image
|
|
|
|
# Usage tips
|
|
- Infra type: CPU
|
|
|
|
# Source code
|
|
```
|
|
class AutoContrast:
|
|
|
|
@classmethod
|
|
def INPUT_TYPES(s):
|
|
return {'required': {'IMAGE': ('IMAGE',), 'cutoff': ('FLOAT', {'default': 2, 'min': 0, 'max': 100, 'step': 0.01}), 'min_value': ('INT', {'default': -1, 'min': -1, 'max': 255, 'step': 1})}}
|
|
RETURN_TYPES = ('IMAGE',)
|
|
FUNCTION = 'process_image'
|
|
CATEGORY = 'image/postprocessing'
|
|
|
|
def process_image(self, IMAGE, cutoff, min_value):
|
|
cimg = conv_tensor_pil(IMAGE)
|
|
if min_value >= 0:
|
|
return conv_pil_tensor(ImageOps.autocontrast(cimg, cutoff=cutoff, ignore=min_value))
|
|
else:
|
|
return conv_pil_tensor(ImageOps.autocontrast(cimg, cutoff=cutoff))
|
|
``` |