Compression Artifact Analysis
JPEG Forensics
Analyzes JPEG compression artifacts to estimate quality levels and detect re-compression patterns that indicate image manipulation or AI generation.
Frequently Asked Questions
How is JPEG quality estimated?
By analyzing DCT coefficients in 8x8 blocks and comparing quantization patterns. Different JPEG quality settings produce distinctive coefficient distributions.
What are blocking artifacts?
JPEG compresses in 8x8 pixel blocks. Heavy compression creates visible block boundaries. AI output often lacks these natural JPEG artifacts or has unusual patterns.
Can you detect double compression?
Yes, re-saving a JPEG creates distinctive dual patterns in DCT histograms. This helps identify manipulated images that were saved multiple times.
Why do AI images have different compression patterns?
AI generates images at the pixel level without camera-like compression. When saved as JPEG, the compression patterns differ from camera-originated images.
Does PNG format bypass this detection?
PNG uses lossless compression, so this specific method doesn't apply. However, other detection methods work on PNG files effectively.
What is quantization noise analysis?
JPEG quantization introduces specific noise patterns. Camera sensors add their own noise on top. AI images lack this layered noise characteristic.
How does multiple re-saves affect detection?
Each re-save adds compression artifacts. Real photos typically show 1-2 compression generations; AI shared on social media may show many more.
What JPEG quality range is most informative?
Quality 70-90% provides the best balance. Higher quality has fewer artifacts to analyze; lower quality obscures original characteristics with heavy compression.
Is this effective against all AI models?
Yes, regardless of the AI model used, the output format determines compression patterns. All AI images show non-camera-like compression when saved as JPEG.
Why is the weight only 7%?
Compression patterns can be altered by format conversion and social media processing. This makes them less reliable than content-based methods, hence the lower weight.
相关方法
机器学习检测
我们的机器学习检测使用最先进的Transformer模型，在数百万张图像上训练，以区分真实照片和AI生成内容。
PRNU分析
光响应非均匀性(PRNU)检测来自制造缺陷的独特相机传感器指纹。AI图像无法复制这些真实的传感器签名。
频率分析
频域分析检查图像中高频和低频分量的分布。AI生成的图像通常缺乏真实照片中存在的自然高频噪声。
梯度分析
使用Sobel、Canny和Laplacian算子分析边缘模式和纹理特征。AI图像通常具有不自然平滑或均匀的梯度。
噪声模式
真实照片包含来自相机传感器的独特噪声模式，这些模式在图像中有所不同。AI生成的图像具有不自然的均匀噪声分布。
元数据分析
图像元数据包含关于其来源的重要线索。我们分析EXIF数据、软件签名和其他嵌入信息以识别AI生成工具。
GAN指纹
检测GAN特有的伪影，如棋盘格图案、色带和生成对抗网络特有的频谱异常。
纹理分析
局部二值模式分析用于检测AI生成图像中常见的纹理异常。测量均匀性、熵和同质性。
人体解剖检测
AI图像生成器经常创建人类立即识别为错误的解剖错误。我们使用计算机视觉来检测这些明显的错误。
C2PA验证
C2PA是通过加密签名跟踪数字内容来源和历史的行业标准。
Semantic Inconsistency Detection
Detects logical inconsistencies like incorrect shadows, impossible perspectives, distorted reflections, and violations of physical laws that AI often produces.
Human Biometric Analysis
Uses MediaPipe to analyze human anatomy for incorrect finger counts, asymmetric eyes, unnatural skin texture, and other anatomical anomalies common in AI-generated faces.
Lighting Physics Validation
Validates light source consistency, shadow direction physics, specular highlight accuracy, and color temperature uniformity across the image.
Edge Sharpness Analysis
Analyzes sharpness distribution across the image and validates depth-of-field consistency. AI often produces unnaturally uniform sharpness.
Statistical Pattern Analysis
Analyzes statistical properties including Shannon entropy, histogram patterns, and Benford's Law compliance to detect synthetic image characteristics.
Chromatic Aberration Analysis
Detects the absence of chromatic aberration (color fringing) that real camera lenses produce. AI images lack these optical artifacts.
Micro-Texture Analysis
Analyzes microscopic texture patterns for repetition, uniformity, and unnatural randomness that AI generators often exhibit.
Color Palette Analysis
Analyzes color distribution including saturation levels, color diversity, and white balance consistency. AI images often have oversaturated colors.