ELA (Error Level Analysis)
Definition
A passive image forensics technique that re-saves the input image at a chosen JPEG quality and amplifies the per-pixel absolute difference to reveal regions at different compression equilibria.
- Technique type
- Passive image forensics
- Method
- Re-save at set JPEG quality, diff pixels
- Reveals
- Regions at different compression equilibria
- Evidence weight
- Indicative, not conclusive
Common questions
Why does ELA only work reliably on JPEG images?+
The technique depends on JPEG's lossy quantisation process reaching a stable equilibrium after repeated saves at the same quality; a region spliced in from a different source or a different original quality has not reached that equilibrium and shows a different error level, a mechanism that does not exist in lossless formats.
Can a genuine, unedited image show ELA artefacts that look like manipulation?+
Yes, sharp edges, saturated colours, and areas that were resaved multiple times can all show elevated error levels without any splicing, which is why ELA results are treated as a lead for further investigation rather than standalone proof of tampering.
Related terms
- Compression Equilibrium
- The state of an image region after sufficient JPEG compression cycles at a given quality: further compression at the same quality produces...
- ELA Amplification Factor
- The multiplier applied to the per-pixel difference before display. A higher factor reveals subtle differences but also amplifies noise, making regions look...
- False Positive (ELA)
- An ELA anomaly in a region that was not manipulated, typically caused by high-frequency texture, JPEG artefacts in the original, or camera...
- High-Frequency Detail
- Fine textures, sharp edges, and small text that JPEG compresses poorly. These regions inherently show high ELA values in authentic, unmodified images...
- Saturation
- In ELA context, the condition where a region's error level has reached near-zero, indicating it has already been compressed to near-equilibrium. A...