diff --git a/src/image-size/index.ts b/src/image-size/index.ts
new file mode 100644
index 00000000..717a7df2
--- /dev/null
+++ b/src/image-size/index.ts
@@ -0,0 +1,412 @@
+import * as fs from 'fs';
+import { ReadStream } from 'fs';
+
+export interface ImageSize {
+    width: number;
+    height: number;
+}
+
+export enum ImageType {
+    JPEG = 'jpeg',
+    PNG = 'png',
+    BMP = 'bmp',
+    GIF = 'gif',
+    WEBP = 'webp',
+    UNKNOWN = 'unknown',
+}
+
+interface ImageParser {
+    readonly type: ImageType;
+    canParse(buffer: Buffer): boolean;
+    parseSize(stream: ReadStream): Promise<ImageSize | undefined>;
+}
+
+// 魔术匹配
+function matchMagic(buffer: Buffer, magic: number[], offset = 0): boolean {
+    if (buffer.length < offset + magic.length) return false;
+
+    for (let i = 0; i < magic.length; i++) {
+        if (buffer[offset + i] !== magic[i]) return false;
+    }
+    return true;
+}
+
+// PNG解析器
+class PngParser implements ImageParser {
+    readonly type = ImageType.PNG;
+    // PNG 魔术头：89 50 4E 47 0D 0A 1A 0A
+    private readonly PNG_SIGNATURE = [0x89, 0x50, 0x4E, 0x47, 0x0D, 0x0A, 0x1A, 0x0A];
+
+    canParse(buffer: Buffer): boolean {
+        return matchMagic(buffer, this.PNG_SIGNATURE);
+    }
+
+    async parseSize(stream: ReadStream): Promise<ImageSize | undefined> {
+        return new Promise((resolve, reject) => {
+            stream.once('error', reject);
+            stream.once('readable', () => {
+                const buf = stream.read(24) as Buffer;
+                if (!buf || buf.length < 24) return resolve(undefined);
+                if (this.canParse(buf)) {
+                    const width = buf.readUInt32BE(16);
+                    const height = buf.readUInt32BE(20);
+                    resolve({ width, height });
+                } else {
+                    resolve(undefined);
+                }
+            });
+        });
+    }
+}
+
+// JPEG解析器
+class JpegParser implements ImageParser {
+    readonly type = ImageType.JPEG;
+    // JPEG 魔术头：FF D8
+    private readonly JPEG_SIGNATURE = [0xFF, 0xD8];
+
+    // JPEG标记常量
+    private readonly SOF_MARKERS = {
+        SOF0: 0xC0, // 基线DCT
+        SOF1: 0xC1, // 扩展顺序DCT
+        SOF2: 0xC2, // 渐进式DCT
+        SOF3: 0xC3, // 无损
+    } as const;
+
+    // 非SOF标记
+    private readonly NON_SOF_MARKERS: number[] = [
+        0xC4, // DHT
+        0xC8, // JPEG扩展
+        0xCC, // DAC
+    ] as const;
+
+    canParse(buffer: Buffer): boolean {
+        return matchMagic(buffer, this.JPEG_SIGNATURE);
+    }
+
+    isSOFMarker(marker: number): boolean {
+        return (
+            marker === this.SOF_MARKERS.SOF0 ||
+            marker === this.SOF_MARKERS.SOF1 ||
+            marker === this.SOF_MARKERS.SOF2 ||
+            marker === this.SOF_MARKERS.SOF3
+        );
+    }
+
+    isNonSOFMarker(marker: number): boolean {
+        return this.NON_SOF_MARKERS.includes(marker);
+    }
+
+    async parseSize(stream: ReadStream): Promise<ImageSize | undefined> {
+        return new Promise<ImageSize | undefined>((resolve, reject) => {
+            const BUFFER_SIZE = 1024; // 读取块大小，可以根据需要调整
+            let buffer = Buffer.alloc(0);
+            let offset = 0;
+            let found = false;
+
+            // 处理错误
+            stream.on('error', (err) => {
+                stream.destroy();
+                reject(err);
+            });
+
+            // 处理数据块
+            stream.on('data', (chunk: Buffer | string) => {
+                // 追加新数据到缓冲区
+                const chunkBuffer = Buffer.isBuffer(chunk) ? chunk : Buffer.from(chunk);
+                buffer = Buffer.concat([buffer.subarray(offset), chunkBuffer]);
+                offset = 0;
+
+                // 保持缓冲区在合理大小内，只保留最后的部分用于跨块匹配
+                const bufferSize = buffer.length;
+                const MIN_REQUIRED_BYTES = 10; // SOF段最低字节数
+
+                // 从JPEG头部后开始扫描
+                while (offset < bufferSize - MIN_REQUIRED_BYTES) {
+                    // 寻找FF标记
+                    if (buffer[offset] === 0xFF && buffer[offset + 1]! >= 0xC0 && buffer[offset + 1]! <= 0xCF) {
+                        const marker = buffer[offset + 1];
+                        if (!marker) break;
+                        // 跳过非SOF标记
+                        if (this.isNonSOFMarker(marker)) {
+                            offset += 2;
+                            continue;
+                        }
+
+                        // 处理SOF标记 (包含尺寸信息)
+                        if (this.isSOFMarker(marker)) {
+                            // 确保缓冲区中有足够数据读取尺寸
+                            if (offset + 9 < bufferSize) {
+                                // 解析尺寸: FF XX YY YY PP HH HH WW WW ...
+                                // XX = 标记, YY YY = 段长度, PP = 精度, HH HH = 高, WW WW = 宽
+                                const height = buffer.readUInt16BE(offset + 5);
+                                const width = buffer.readUInt16BE(offset + 7);
+
+                                found = true;
+                                stream.destroy();
+                                resolve({ width, height });
+                                return;
+                            } else {
+                                // 如果缓冲区内数据不够，保留当前位置等待更多数据
+                                break;
+                            }
+                        }
+                    }
+
+                    offset++;
+                }
+
+                // 缓冲区管理: 如果处理了许多数据但没找到标记，
+                // 保留最后N字节用于跨块匹配，丢弃之前的数据
+                if (offset > BUFFER_SIZE) {
+                    const KEEP_BYTES = 20; // 保留足够数据以处理跨块边界的情况
+                    if (offset > KEEP_BYTES) {
+                        buffer = buffer.slice(offset - KEEP_BYTES);
+                        offset = KEEP_BYTES;
+                    }
+                }
+            });
+
+            // 处理流结束
+            stream.on('end', () => {
+                if (!found) {
+                    resolve(undefined);
+                }
+            });
+        });
+    }
+}
+
+// BMP解析器
+class BmpParser implements ImageParser {
+    readonly type = ImageType.BMP;
+    // BMP 魔术头：42 4D (BM)
+    private readonly BMP_SIGNATURE = [0x42, 0x4D];
+
+    canParse(buffer: Buffer): boolean {
+        return matchMagic(buffer, this.BMP_SIGNATURE);
+    }
+
+    async parseSize(stream: ReadStream): Promise<ImageSize | undefined> {
+        return new Promise((resolve, reject) => {
+            stream.once('error', reject);
+            stream.once('readable', () => {
+                const buf = stream.read(26) as Buffer;
+                if (!buf || buf.length < 26) return resolve(undefined);
+                if (this.canParse(buf)) {
+                    const width = buf.readUInt32LE(18);
+                    const height = buf.readUInt32LE(22);
+                    resolve({ width, height });
+                } else {
+                    resolve(undefined);
+                }
+            });
+        });
+    }
+}
+
+// GIF解析器
+class GifParser implements ImageParser {
+    readonly type = ImageType.GIF;
+    // GIF87a 魔术头：47 49 46 38 37 61
+    private readonly GIF87A_SIGNATURE = [0x47, 0x49, 0x46, 0x38, 0x37, 0x61];
+    // GIF89a 魔术头：47 49 46 38 39 61
+    private readonly GIF89A_SIGNATURE = [0x47, 0x49, 0x46, 0x38, 0x39, 0x61];
+
+    canParse(buffer: Buffer): boolean {
+        return (
+            matchMagic(buffer, this.GIF87A_SIGNATURE) ||
+            matchMagic(buffer, this.GIF89A_SIGNATURE)
+        );
+    }
+
+    async parseSize(stream: ReadStream): Promise<ImageSize | undefined> {
+        return new Promise((resolve, reject) => {
+            stream.once('error', reject);
+            stream.once('readable', () => {
+                const buf = stream.read(10) as Buffer;
+                if (!buf || buf.length < 10) return resolve(undefined);
+                if (this.canParse(buf)) {
+                    const width = buf.readUInt16LE(6);
+                    const height = buf.readUInt16LE(8);
+                    resolve({ width, height });
+                } else {
+                    resolve(undefined);
+                }
+            });
+        });
+    }
+}
+
+// WEBP解析器 - 完整支持VP8, VP8L, VP8X格式
+class WebpParser implements ImageParser {
+    readonly type = ImageType.WEBP;
+    // WEBP RIFF 头：52 49 46 46 (RIFF)
+    private readonly RIFF_SIGNATURE = [0x52, 0x49, 0x46, 0x46];
+    // WEBP 魔术头：57 45 42 50 (WEBP)
+    private readonly WEBP_SIGNATURE = [0x57, 0x45, 0x42, 0x50];
+
+    // WEBP 块头
+    private readonly CHUNK_VP8 = [0x56, 0x50, 0x38, 0x20];  // "VP8 "
+    private readonly CHUNK_VP8L = [0x56, 0x50, 0x38, 0x4C]; // "VP8L"
+    private readonly CHUNK_VP8X = [0x56, 0x50, 0x38, 0x58]; // "VP8X"
+
+    canParse(buffer: Buffer): boolean {
+        return (
+            buffer.length >= 12 &&
+            matchMagic(buffer, this.RIFF_SIGNATURE, 0) &&
+            matchMagic(buffer, this.WEBP_SIGNATURE, 8)
+        );
+    }
+
+    isChunkType(buffer: Buffer, offset: number, chunkType: number[]): boolean {
+        return buffer.length >= offset + 4 && matchMagic(buffer, chunkType, offset);
+    }
+
+    async parseSize(stream: ReadStream): Promise<ImageSize | undefined> {
+        return new Promise((resolve, reject) => {
+            // 需要读取足够的字节来检测所有三种格式
+            const MAX_HEADER_SIZE = 32;
+            let totalBytes = 0;
+            let buffer = Buffer.alloc(0);
+
+            stream.on('error', reject);
+
+            stream.on('data', (chunk: Buffer | string) => {
+                const chunkBuffer = Buffer.isBuffer(chunk) ? chunk : Buffer.from(chunk);
+                buffer = Buffer.concat([buffer, chunkBuffer]);
+                totalBytes += chunk.length;
+
+                // 检查是否有足够的字节进行格式检测
+                if (totalBytes >= MAX_HEADER_SIZE) {
+                    stream.destroy();
+
+                    // 检查基本的WEBP签名
+                    if (!this.canParse(buffer)) {
+                        return resolve(undefined);
+                    }
+
+                    // 检查chunk头部，位于字节12-15
+                    if (this.isChunkType(buffer, 12, this.CHUNK_VP8)) {
+                        // VP8格式 - 标准WebP
+                        // 宽度和高度在帧头中
+                        const width = buffer.readUInt16LE(26) & 0x3FFF;
+                        const height = buffer.readUInt16LE(28) & 0x3FFF;
+                        return resolve({ width, height });
+
+                    } else if (this.isChunkType(buffer, 12, this.CHUNK_VP8L)) {
+                        // VP8L格式 - 无损WebP
+                        // 1字节标记后是14位宽度和14位高度
+                        const bits = buffer.readUInt32LE(21);
+                        const width = 1 + (bits & 0x3FFF);
+                        const height = 1 + ((bits >> 14) & 0x3FFF);
+                        return resolve({ width, height });
+
+                    } else if (this.isChunkType(buffer, 12, this.CHUNK_VP8X)) {
+                        // VP8X格式 - 扩展WebP
+                        // 24位宽度和高度(减去1)
+                        if (!buffer[24] || !buffer[25] || !buffer[26] || !buffer[27] || !buffer[28] || !buffer[29]) {
+                            return resolve(undefined);
+                        }
+                        const width = 1 + ((buffer[24] | (buffer[25] << 8) | (buffer[26] << 16)) & 0xFFFFFF);
+                        const height = 1 + ((buffer[27] | (buffer[28] << 8) | (buffer[29] << 16)) & 0xFFFFFF);
+                        return resolve({ width, height });
+                    } else {
+                        // 未知的WebP子格式
+                        return resolve(undefined);
+                    }
+                }
+            });
+
+            stream.on('end', () => {
+                // 如果没有读到足够的字节
+                if (totalBytes < MAX_HEADER_SIZE) {
+                    resolve(undefined);
+                }
+            });
+        });
+    }
+}
+
+const parsers: ReadonlyArray<ImageParser> = [
+    new PngParser(),
+    new JpegParser(),
+    new BmpParser(),
+    new GifParser(),
+    new WebpParser(),
+];
+
+export async function detectImageType(filePath: string): Promise<ImageType> {
+    return new Promise((resolve, reject) => {
+        const stream = fs.createReadStream(filePath, {
+            highWaterMark: 64, // 优化读取buffer大小
+            start: 0,
+            end: 63
+        });
+
+        let buffer: Buffer | null = null;
+
+        stream.once('error', (err) => {
+            stream.destroy();
+            reject(err);
+        });
+
+        stream.once('readable', () => {
+            buffer = stream.read(64) as Buffer;
+            stream.destroy();
+
+            if (!buffer) {
+                return resolve(ImageType.UNKNOWN);
+            }
+
+            for (const parser of parsers) {
+                if (parser.canParse(buffer)) {
+                    return resolve(parser.type);
+                }
+            }
+
+            resolve(ImageType.UNKNOWN);
+        });
+
+        stream.once('end', () => {
+            if (!buffer) {
+                resolve(ImageType.UNKNOWN);
+            }
+        });
+    });
+}
+
+export async function imageSizeFromFile(filePath: string): Promise<ImageSize | undefined> {
+    try {
+        // 先检测类型
+        const type = await detectImageType(filePath);
+        const parser = parsers.find(p => p.type === type);
+        if (!parser) return undefined;
+
+        // 用流式方式解析尺寸
+        const stream = fs.createReadStream(filePath);
+        try {
+            return await parser.parseSize(stream);
+        } catch (err) {
+            console.error(`解析图片尺寸出错: ${err}`);
+            return undefined;
+        } finally {
+            if (!stream.destroyed) {
+                stream.destroy();
+            }
+        }
+    } catch (err) {
+        console.error(`检测图片类型出错: ${err}`);
+        return undefined;
+    }
+}
+
+export async function imageSizeFallBack(
+    filePath: string,
+    fallback: ImageSize = {
+        width: 1024,
+        height: 1024,
+    }
+): Promise<ImageSize> {
+    return await imageSizeFromFile(filePath) ?? fallback;
+}
\ No newline at end of file