SIFT (Scale-Invariant Feature Transform)
Definition
A keypoint descriptor algorithm that detects interest points and computes 128-dimensional descriptors invariant to scale and rotation. In forgery detection, matching SIFT descriptors within a single image identifies regions with identical local structure that would not be present in an authentic photograph.
- Output
- 128-dimensional keypoint descriptors
- Invariance
- Scale and rotation
- Forgery detection use
- Matching descriptors within a single image to find copy-move regions
- Signal of tampering
- Identical local structure not expected in an authentic photograph
Common questions
How does matching SIFT descriptors within one image reveal a copy-move forgery?+
In an authentic photograph, no two regions normally share near-identical local texture and structure by chance. If SIFT finds a cluster of keypoints in one region whose descriptors closely match keypoints in another region of the same image, that pairing indicates one area was duplicated from the other, even if the copied region was rotated, resized, or moved.
Can SIFT-based detection be defeated by simple edits like blurring or JPEG recompression?+
Heavy blurring, strong recompression, or added noise can degrade descriptor matching and reduce detection sensitivity, since SIFT relies on distinct local gradient structure that these operations smooth away, so examiners often combine SIFT matching with other forgery detection methods rather than relying on it alone against a sophisticated tamperer.
Related terms
- Affine Transform Estimation
- The recovery of the geometric transformation (rotation, scaling, shear) relating a copied region to its source, using matched keypoint pairs. Geometrically consistent...
- Block-Matching
- A copy-move detection strategy that divides the image into overlapping fixed-size blocks, computes a compact feature vector per block, sorts vectors, and...
- Chromatic Aberration
- Colour fringing at high-contrast edges caused by differential refraction of light wavelengths by a lens. The pattern is lens-specific and inconsistent across...
- Copy-Move Forgery
- An intra-image manipulation in which a patch copied from one location within the image is pasted over another location in the same...
- Localisation Map
- The output of a forgery detection algorithm that marks, at pixel or block resolution, which regions of the image are identified as...
- Noiseprint
- A CNN-based camera-model fingerprint extractor by Cozzolino and Verdoliva. Applied to deepfakes, it reveals inconsistency between the camera fingerprint in the genuine...
- Passive Forgery Detection
- Detection that operates on the image data alone, without any pre-embedded watermark or signature. Distinguished from active methods such as digital watermarking...
- Splicing
- An inter-image manipulation in which content from one or more external images is inserted into a target image. Splicing introduces cross-boundary inconsistencies...