Similar Image Finder & Visual Duplicate Detector

Discover how the browser-based Universal Similar Image Finder & Visual Duplicate Detector Studio on RiazHub.com leverages 64-bit difference hashing (dHash), average hashing (aHash), and block-mean RGB chromatic moments to detect resized, recompressed, watermarked, and near-identical image copies across large photo libraries. Explore dual scan topologies (All-to-All deduplication vs. 1-to-Many reference needle search), synchronized side-by-side zoom stages, false-color delta difference heatmaps, 60 FPS swipe dividers, intelligent auto-selection rules (megapixels, byte weight, and timestamps), and zero-dependency in-memory PKZIP library archiving all executed 100% client-side with complete privacy and zero server uploads.

Universal Similar Image Finder v2.5 Pro Studio

Detect exact duplicates, identify resized, re-compressed, or watermarked copies via 64-bit perceptual hashing, inspect visual differences, and clean your photo library in real time.

📁 Library Ingestion

0 Images

0.0 MB total size

🔍 Duplicate Clusters

0 Groups

0 Redundant files flagged

⚡ Recoverable Space

0.0 MB

0% storage gain

🛡️ Processing Engine

100% In-Browser

64-bit dHash & aHash Client-Side

⚡ Presets:

🎯 Resized & Compressed

🔒 Exact Duplicates

📸 Burst Shot Clusterer

🎨 Color Variant Finder

🖼️

Drop images or folder here

Supports JPG, PNG, WEBP, AVIF, BMP, GIF (Up to 500+)

📁 Select Image Files 🗂️ Select Entire Folder (webkitdirectory) 🧪 Load Sample Duplicate Set

Scan Topology All-to-All

🗂️

All-to-All Deduplication Scan entire library for pairwise clusters

🎯

1-to-Many Needle Search Match reference target photo against library

🎯 Reference Image Target (Needle)

Drop or select a target image to match across your library:

Select Reference Image

Perceptual Algorithm Engine

dHash (Gradient) aHash (Mean) Dual Composite

Hamming Distance Tolerance: 4 bits (≥93.8% Match)

0 (Identical) 4 (Recompressed) 12 (Burst Shot) 16 (Loose)

Color Moment Weight Bias: 20% Weight

0% (Geometry only) 50% (Balanced) 100% (Strict Palette)

🪄 Intelligent Auto-Select Rules 1-Click

📉 Keep Highest Resolution Flag lower megapixel copies 💾 Keep Smallest File Flag heavier redundant files ⏳ Keep Oldest Original Flag newer edits & copies ⏱️ Keep Newest Edit Flag older original snaps

⚡ Scan for Similar Images

Calculating Perceptual Hashes... 0%

🔄 Reset Params 🗑️ Clear Library

🗂️ Similar Clusters Grid 0 🔍 Side-by-Side Zoom 🔥 Delta Heatmap ↔️ Split Swipe Slider 📊 Audit Table

Select All Redundant Invert Selection Deselect All

Sort: Highest Similarity Largest File Size Most Matches in Group

📂

No Scans Performed Yet

Load your photos or click "Load Sample Duplicate Set" to immediately test the perceptual hashing engine.

Compare Pair: vs Reset Zoom (100%)

Image A: No selection

Image B: No selection

💡 Tip: Scroll with your mouse wheel over either viewport to zoom synchronously; click and drag to pan both views simultaneously.

Heatmap Delta: Select pair to generate

Threshold: 25

🔥

Select a cluster pair in Tab 1 or dropdown above to compute per-pixel difference heatmap

Magenta / Pink regions highlight pixel differences, watermarks, or crop offsets.

Swipe Compare: Select a duplicate pair

↔

Drag the vertical bar left and right to inspect subtle recompression artifacts or geometry differences.

Flag	Filename	Cluster #	Match %	Resolution	File Size	Recommendation	Action
No audit records generated. Run a scan first.

📦 Download Unique Clean ZIP 🗑️ Purge Selected Duplicates 📊 Export CSV Audit 📋 Copy Duplicate File List

Perceptual Hashing Mathematics, Hamming Distance & Clean Deduplication Guide

1. Cryptographic Hashes (SHA-256) vs. Perceptual Hashes (dHash / aHash) ▼

Cryptographic hashing algorithms (such as MD5, SHA-1, or SHA-256) are designed to exhibit an extreme avalanche effect: changing a single bit in an image file results in an entirely different hash string. Therefore, if an image is saved as a WebP instead of a JPG, resized by 1 pixel, or slightly recompressed, cryptographic checksums will fail completely to detect the duplicate.

In contrast, Perceptual Hashing (pHash, dHash, aHash) translates visual frequencies and gradient structures into compact 64-bit fingerprints. Images that look visually similar to human eyes produce identical or near-identical binary bit sequences regardless of format, metadata changes, or compression.

2. Mathematical Formula: 64-bit Difference Hash (dHash) & Gradient Transitions ▼

The Difference Hash (dHash) algorithm analyzes gradient flows between adjacent pixels:

1. Bilinear downsample candidate image to 9 x 8 grayscale matrix (72 pixels total).
2. For each row y ∈ [0, 7] and column x ∈ [0, 7]:
P(x, y) > P(x + 1, y) ⟹ bit = 1, otherwise bit = 0
3. Produces a compact 64-bit unsigned binary vector.

Because it evaluates relative brightness between neighboring pixels rather than absolute luminance, dHash is invariant to global contrast boosts, brightness adjustments, and aspect ratio modifications.

3. Bitwise XOR Hamming Distance & Popcount In-Memory Comparison ▼

Distance between two 64-bit image fingerprints $H_A$ and $H_B$ is evaluated via the bitwise XOR population count:

D_H(H_A, H_B) = popcount(H_A ⊕ H_B) = ∑ (H_A[i] ⊕ H_B[i]) for i=0..63
Similarity % = (1 - D_H / 64) × 100%

Hamming Distance Interpretation:

• Distance 0 (100%): Exact identical perceptual match.
• Distance 1 – 4 (94% – 98%): Minor compression artifacts, slight watermarking, format conversion (PNG to WebP).
• Distance 5 – 9 (86% – 92%): Burst camera shots, micro-movements, slight color grading.
• Distance 10+ (<84%): Distinct pictures with loosely similar layout composition.

4. 100% Client-Side In-Browser Privacy Guarantee ▼

All image decoding, pixel matrix downsampling, perceptual hash generation, pairwise distance evaluations, visual heatmap comparisons, and clean PKZIP packaging occur 100% within your local browser sandbox via native HTML5 Canvas, typed arrays, and JavaScript Web APIs. No photos, EXIF data, or filenames are ever transmitted to external cloud servers or stored in any database.

✓ Notification