The Universal Image Duplicate Finder & Visual Similarity Deduplication Studio by RiazHub is an advanced, 100% client-side digital utility designed to scan, detect, and eliminate redundant image files directly inside your browser with zero server uploads. Powered by a dual-tier matching engine, it pairs native cryptographic SHA-256 byte hashing (for instant detection of exact clones, redundant downloads, and byte-for-byte copies) with 64-bit perceptual fingerprinting algorithms—including dHash (difference gradient matching), aHash (luminance mean thresholding), and DCT-based pHash (discrete cosine transform frequency domain analysis) to identify near-duplicates, compressed copies, resized web variants, and rapid camera burst shots.
Features include an interactive similarity threshold slider (70% to 100% Hamming tolerance), automated disjoint-set cluster grouping, and an intelligent “Keep Best Shot” recommendation engine that evaluates candidates by pixel resolution ($W \times H$), file size, EXIF timestamps, or edge-contrast Laplacian sharpness. Users can verify candidates using an interactive side-by-side split-view diff slider, inspect pixel subtraction variations in a difference heatmap, and selectively download curated unique libraries or marked duplicates via an in-memory zero-dependency PKZIP generator alongside comprehensive CSV and JSON audit reports.
Universal Image Duplicate Finder & Similarity Studio
Scan albums for exact clones and visually similar photos, compare differences side-by-side, auto-select low-res copies, and reclaim storage space in real time. Zero server uploads — completely private.
| Thumb | File Name | Dimensions | File Size | Sharpness | Status | Action |
|---|---|---|---|---|---|---|
| No photos scanned yet. | ||||||
10110010...). Hamming distance measures the exact count of bit positions that differ between two fingerprints. If two photos have a Hamming distance of 0, their visual structure is identical (100% similarity). A distance of 6 out of 64 corresponds to $\approx 90.6\%$ similarity. In this studio:
- dHash (Difference Hash): Tracks relative horizontal luminance gradients across a 9×8 downsampled matrix. Extremely fast and resilient to brightness shifts.
- aHash (Average Hash): Computes the mean luminance across an 8×8 grid and thresholds every pixel against that mean.
- pHash (Perceptual DCT Hash): Applies a 2D Discrete Cosine Transform (DCT) to downscaled 32×32 luminance data to extract the 8×8 lowest spatial frequency components, making it exceptionally resistant to rotation, blur, and aggressive lossy compression.
- Highest Pixel Resolution: Computes $W \times H$ megapixels. Preserves original full-sized master files while marking downscaled web variants for cleanup.
- Largest File Size: Identifies uncompressed or minimally compressed versions with maximal pixel color fidelity.
- Oldest Date / Origin: Favors the earliest timestamped or originally indexed file.
- Sharpest / Least Blurry: Runs an edge-contrast Laplacian variance filter on a normalized 256×256 canvas to automatically isolate the crispest photo in rapid camera bursts.
crypto.subtle, 64-bit perceptual hashing, connected-components clustering, visual diff rendering, and in-memory ZIP package creation—occurs strictly within your device's browser memory. No photo data, thumbnails, filenames, or metadata are ever transmitted over the network to any server.