Noise Analysis Scanner

Reveals manipulation traces by extracting the noise pattern from an image. Edited regions show inconsistent noise levels — airbrushing, warping, and clone stamping leave visible marks.

500+ images analyzed

Drag & drop a file here, or

JPG PNG WebP GIF BMP AVIF Max 50MB • Processed in browser

Reverse Median Filter

Subtracts a median-filtered version from the original to isolate the high-frequency noise residual.

Block Analysis

Divides the noise map into blocks and compares variance across regions to detect localized manipulation.

100% Private

All analysis runs locally in your browser via Web Worker. Nothing is uploaded to any server.

How noise analysis works

Every digital photo contains sensor noise — a subtle random pattern created during capture. Noise analysis extracts this pattern by applying a median filter (which preserves edges but removes noise) and subtracting the result from the original image. What remains is the noise residual. In an unedited photo, this noise is distributed uniformly. Manipulation tools like airbrush, clone stamp, warp, and content-aware fill disrupt the noise pattern, creating visible inconsistencies. For compression-based detection, combine with the ELA Scanner. For copy-paste forgery, use Clone Detection.

Understanding the controls

The kernel size controls the median filter window. A 3×3 kernel captures fine noise detail but may include edge artifacts. A 5×5 or 7×7 kernel produces a smoother noise map that better highlights large retouched areas but loses fine-grained information. Amplification (1–50×) magnifies the noise residual for visibility. RGB mode shows per-channel noise in color; luminance mode converts to grayscale for easier pattern spotting. Histogram equalization stretches the contrast to make subtle differences more visible. Use the Re-run button to adjust parameters without re-uploading.

Who uses noise analysis

Photo editors and retouchers use noise analysis to quality-check their own work and ensure seamless blending. Forensic analysts detect airbrushing and beauty retouching in evidence photos. Academic integrity reviewers check scientific images in journal submissions for undisclosed manipulation. Insurance adjusters examine claim photos for suspicious smoothing or patching. For metadata-level verification, combine with the Authenticity Checker. For a comprehensive forensic workflow, visit the photo forensics hub.

Noise analysis vs ELA

ELA detects compression inconsistencies — it works best on JPEG images that have been through at least one compression cycle. Noise analysis detects texture-level manipulation regardless of format. ELA catches pasted regions with different compression history; noise analysis catches retouching, smoothing, and local noise reduction that ELA may miss entirely. These tools complement each other. For splice detection via JPEG quality levels, add the JPEG Ghost Scanner. For hidden data analysis, try the Stego Scanner.

Noise analysis is one of the most effective techniques for detecting subtle retouching that other forensic methods miss. Beauty retouching, skin smoothing, and selective noise reduction all suppress the natural sensor noise in affected regions, making them stand out clearly in the noise map. The block-based uniformity analysis automatically flags regions with anomalous noise levels. All processing runs entirely in your browser via a Web Worker — your images are never uploaded. For checking whether an image was AI-generated, use the AI Detector. For thumbnail-based editing detection, see the Thumbnail Scanner. For device-level identification via sensor noise patterns, try the Camera Fingerprint (PRNU) tool. Learn more in our articles on detecting photo edits and image forensics guide.

Frequently Asked Questions

What is noise analysis in image forensics?

Noise analysis extracts the sensor noise pattern from an image by subtracting a median-filtered version from the original. In unedited photos, noise is distributed uniformly. Manipulation tools like airbrush, clone stamp, and content-aware fill disrupt this pattern, creating visible inconsistencies in the noise map.

What kind of manipulation does noise analysis detect?

Noise analysis is particularly effective at detecting airbrushing, beauty retouching, skin smoothing, selective noise reduction, warp/liquify distortions, and clone stamp blending. These tools suppress or alter the natural noise pattern in affected areas. It complements ELA, which focuses on compression artifacts instead.

What do the different kernel sizes do?

The kernel size controls the median filter window. A 3x3 kernel captures fine noise detail and subtle edits. A 5x5 kernel produces a smoother noise map that highlights larger retouched areas. A 7x7 kernel is best for spotting broad manipulations like background replacement. Try different sizes and use the Re-run button to compare.

What does uniform vs inconsistent noise mean?

Uniform noise means the noise level is similar across the entire image, typical of unedited photos. Inconsistent noise means some regions have significantly different noise levels, often indicating localized manipulation. The coefficient of variation measures this difference — higher values suggest more inconsistency.

Can noise analysis detect AI-generated images?

Noise analysis may show unusual patterns in AI-generated images since they lack natural sensor noise. However, for dedicated AI detection, the AI Detector tool uses purpose-built models and is more reliable for this specific task.

Is my image uploaded to a server?

No. All noise analysis processing runs entirely in your browser using a Web Worker. The image never leaves your device. No data is transmitted, stored, or logged.