Interpret

Takes a raw pixel diff and tells you what changed, where, and how much. No ML models — a deterministic pipeline that runs in the same binary as the diff itself.

How it works

1. Pixel diff → binary change mask
2. Morphological close → bridge small gaps
3. Connected components → isolate regions
4. Per-region evidence extraction:
   • Dual-image gradients — edges in both images + spatial correlation to detect structural preservation
   • Color delta distribution — mean, max, and stddev of YIQ distance to separate uniform recolors from patchy texture changes
   • Background distance — how much changed pixels blend with local unchanged pixels in each image
5. Six-label decision tree classifies each region
6. Post-hoc shift detection matches Addition+Deletion pairs

Demo

Different Images

Image 1

Image 2

Moderate visual change detected (1.87% of image, 10 regions). Content changed: 4 regions (bottom, center). Content added: 3 regions (right, bottom, bottom-left). Content removed: 3 regions (bottom, top-left, center).

medium1.87% changed

Regions (10)Hover a region to highlight

bottom·deletion·mixed-region·(670, 977, 199×89) · 0.58%

bottom·content-change·sparse-distributed·(558, 738, 193×173) · 0.34%

bottom·content-change·mixed-region·(726, 910, 208×51) · 0.30%

center·content-change·sparse-distributed·(368, 663, 323×88) · 0.27%

bottom·content-change·sparse-distributed·(432, 937, 50×151) · 0.11%

right·addition·sparse-distributed·(1004, 626, 71×130) · 0.08%

top-left·deletion·edge-dominated·(343, 102, 39×126) · 0.07%

center·deletion·sparse-distributed·(505, 717, 49×88) · 0.05%

bottom·addition·sparse-distributed·(694, 1085, 80×33) · 0.04%

bottom-left·addition·edge-dominated·(253, 1191, 89×37) · 0.03%

Show raw JSON

import { interpret } from "@blazediff/core-native";
 
const result = await interpret("fixtures/3a.png", "fixtures/3b.png");
 
for (const region of result.regions) {
  console.log(`${region.position}: ${region.changeType} (${region.percentage.toFixed(2)}%)`);
}

Change types