MAIkit

AI Image OperatorsAI 图像算子

Run classic computer-vision operators on an image: Sobel edges, Otsu thresholding, histogram equalisation, Kuwahara stylisation.

在本地对图片运行经典计算机视觉算子:Sobel 边缘检测、Otsu 自动阈值、直方图均衡、马赛克与 Kuwahara 油画风格。

What is AI Image Operators?AI 图像算子 是什么?

A set of classic computer-vision operators implemented from scratch in JavaScript, so you can see what each algorithm actually does to an image. Sobel edge detection maps intensity gradients to dark outlines, Otsu finds the optimal threshold between foreground and background automatically, histogram equalisation stretches contrast, box and Gaussian convolution blur, mosaicing averages blocks, and a Kuwahara-style filter paints the image like an oil painting. Every operator runs locally on a canvas, so you can chain them and export the result as PNG.

一组用 JavaScript 从零实现的经典计算机视觉算子,让你看清每个算法到底对图像做了什么:Sobel 边缘检测把亮度梯度映射成深色轮廓,Otsu 自动在前景与背景之间选出最优阈值,直方图均衡拉伸对比度,盒式与高斯卷积做模糊,马赛克按块取平均,Kuwahara 风格滤波把图像画成油画质感。所有算子都在本地 canvas 上运行,可以连续叠加并导出 PNG。

How to use ai image operators如何使用 AI 图像算子

  1. Drop in an image (it is scaled to a sensible working size).
  2. Pick an operator and adjust the block size where it applies.
  3. Press Apply to run the operator on the canvas.
  4. Download the result as PNG.
  1. 把图片拖进来(会自动缩放到合适的工作尺寸)。
  2. 选择算子,涉及块大小的可调滑块。
  3. 点"应用",算子会在画布上运行。
  4. 下载结果 PNG。

Common uses常见用途

FAQ常见问题

What does the Otsu operator actually do?Otsu 算子到底做了什么?

Otsu is a threshold-selection method: it scans every possible grey level, computes the variance of the resulting two-class separation, and picks the level that maximises it. The tool prints the threshold it found, so you can see the algorithm decide rather than guessing a value by eye.

Otsu 是一种阈值选择算法:遍历所有灰度级,计算二分类后的类间方差,取方差最大处的灰度作为阈值。工具会把算出的阈值打印出来,让你看到算法在"做决定",而不是靠肉眼猜一个值。

Why does my photo look different after Kuwahara?为什么 Kuwahara 之后图片不一样了?

Kuwahara keeps the most homogeneous square neighbourhood around each pixel, which flattens texture while keeping edges — that is exactly the look of oil paint. A larger block size gives a stronger painting effect and takes longer.

Kuwahara 取每个像素邻域中最均匀的方形区域作为输出,抹平纹理但保留边缘,这正是油画质感。块越大,油画味越强,耗时也越长。

Is this real AI?这算真 AI 吗?

These are the classical operators that deep learning is built on top of, not neural networks. They are included because they are genuinely useful, run instantly and need no model download. For learning how the classic methods work, tracing every step is the point.

这些是深度学习之下的经典算子,不是神经网络。收录它们是因为它们确实有用、瞬时完成、无需下载模型。想搞懂经典方法怎么工作,逐步看每一步才有意义。

Are my images uploaded?图片会上传吗?

No. The image is read with FileReader into a canvas and processed locally.

不会。图片通过 FileReader 读入 canvas,全部本地处理。

Other tools其他工具

🧠

Machine Learning Lab机器学习实验台

Train linear regression, logistic regression and K-means cluster…

在浏览器里直接训练线性回归、逻辑回归与 K-means 聚类,实时画出拟合直线、…

🔬

Text Intelligence文本智能分析

Tokenise text, extract TF-weighted keywords and compare two docu…

对中英文文本分词、去停用词、提取 TF 加权关键词,并用余弦相似度比较两段文本的…

💬

Sentiment Analysis情感分析

Score the sentiment of a review or message with a bilingual posi…

用中英文正负情感词典对评论、留言打分,正确处理"不""没"等否定词,列出命中的正…

🔊

Text to Speech语音合成

Turn text into spoken audio with the browser speech engine, adju…

调用浏览器内置语音引擎朗读文本,可调语速与音调,支持中文、英文、日文与韩文。…

🎤

Speech to Text语音识别

Dictate with your microphone and get live transcribed text using…

用麦克风实时听写,自动转成文字,支持中英文识别,结果可一键复制。…

🧩

Prompt Engineer提示词工程助手

Build structured prompts from role, task, context and constraint…

按角色、任务、背景、约束、输出格式拼出结构化提示词,估算 token 数量与成本…

Sister sites友情链接 MonkeyDev dev tools hub工具总站 FreePDFKit PDF toolsPDF 工具集 ImageToolKit image tools图片工具集 AI 咒语库 AI prompt libraryAI 提示词库 DevTools developer toolkit开发工具箱 CalcKit finance calculators财务计算器 OfficeKit office utilities办公工具箱 DataKit data & spreadsheet tools数据与表格工具 ChartKit diagrams & charts图表与可视化 DocKit markdown & document toolsMarkdown 与文档工具