Files
SnapOtter/apps/docs/zh-CN/tools/image/noise-removal.md
T
SnapOtterandGitHub 4963ab3bbd feat(docs-i18n): translate all documentation into 20 languages
All 181 docs markdown files translated into 20 languages (apps/docs/<locale>/**). Companion to the i18n code PR; admin-merged because the file count exceeds GitHub's per-PR CI trigger limit. Validated by pnpm i18n:check (all surfaces, 0 stale/missing) and a clean all-locale docs build.
2026-07-11 13:52:47 +08:00

80 lines
2.7 KiB
Markdown
Raw Blame History

This file contains ambiguous Unicode characters
This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.
---
description: "基于 AI 的降噪和去颗粒处理,提供多档质量选项。"
i18n_source_hash: f0dfc876e0e0
i18n_provenance: human
i18n_output_hash: 43d042c4f560
---
# 降噪 {#noise-removal}
基于 AI 的降噪和去颗粒处理,提供多档质量选项,使用 Python 边车(SCUNet 模型)。
## API 端点 {#api-endpoint}
`POST /api/v1/tools/image/noise-removal`
**处理方式:** 异步(返回 202,通过 SSE 轮询 `/api/v1/jobs/{jobId}/progress` 获取状态)
**模型包:** `upscale-enhance`5-6 GB
## 参数 {#parameters}
| 参数 | 类型 | 是否必填 | 默认值 | 说明 |
|-----------|------|----------|---------|-------------|
| file | file | 是 | - | 图片文件(multipart |
| tier | string | 否 | `"balanced"` | 质量档位:`quick``balanced``quality``maximum` |
| strength | number | 否 | `50` | 降噪强度(0-100 |
| detailPreservation | number | 否 | `50` | 保留细节的程度(0-100)。值越高保留的纹理越多 |
| colorNoise | number | 否 | `30` | 彩色噪点抑制强度(0-100) |
| format | string | 否 | `"original"` | 输出格式:`original``png``jpeg``webp``avif``jxl` |
| quality | number | 否 | `90` | 输出编码质量(1-100 |
## 请求示例 {#example-request}
```bash
curl -X POST http://localhost:1349/api/v1/tools/image/noise-removal \
-F "file=@noisy-photo.jpg" \
-F 'settings={"tier":"quality","strength":60,"detailPreservation":70,"colorNoise":40}'
```
## 响应 {#response}
### 初始响应(202 Accepted {#initial-response-202-accepted}
```json
{
"jobId": "a1b2c3d4-e5f6-7890-abcd-ef1234567890",
"async": true
}
```
### 进度(位于 `/api/v1/jobs/{jobId}/progress` 的 SSE {#progress-sse-at-api-v1-jobs-jobid-progress}
```
event: progress
data: {"phase":"processing","stage":"Denoising...","percent":65}
```
### 最终结果(通过 SSE {#final-result-via-sse}
```json
{
"phase": "complete",
"percent": 100,
"result": {
"jobId": "a1b2c3d4-e5f6-7890-abcd-ef1234567890",
"downloadUrl": "/api/v1/download/{jobId}/noisy-photo_denoised.jpg",
"originalSize": 500000,
"processedSize": 380000
}
}
```
## 说明 {#notes}
- 需要安装 `upscale-enhance` 模型包(5-6 GB)。
- 质量档位在速度和质量之间权衡:`quick` 最快,仅做基础降噪;`maximum` 采用最彻底的多轮处理方式。
- 对于有纹理的主体(布料、头发、树叶),`detailPreservation` 参数至关重要。较高的值可防止降噪器抹平细微的细节。
-`format` 设为 `"original"` 时,输出格式与输入文件格式一致。
- 通过自动解码支持 HEIC/HEIF、RAW、TGA、PSD、EXR 和 HDR 输入格式。