resemblance-detection · research · also Chunk resemblance detection, Similarity-based base selection

Resemblance detection

Resemblance detection identifies similar, potentially nonidentical data objects as candidates for encoding one relative to another.

Selection: A compact similarity summary narrows the search for a useful base without requiring the target to be a new version of the same named file. REBL first removes duplicate chunks, then searches for resembling chunks and applies delta encoding. [1]

Super-fingerprints: Grouping several features into a hash-table key reduces candidate-matching work. More restrictive matching can miss useful bases; a selected candidate can still compress poorly. Similarity selection is separate from the correctness of reconstructing the target bytes. [1]

Recovery: A stored delta creates a dependency on its selected base. REBL assigns a chunk either as a reference or as an encoded version, rather than repeatedly using encoded versions as new bases. [1]

Relations

related
Finesse, Delta compression, Data deduplication, Content-defined chunking, MinHash

Citations

  1. Kulkarni, Douglis, LaVoie and Tracey. Redundancy Elimination Within Large Collections of Files. USENIX ATC 2004, sections 2-4. (accessed 2026-11-22)
Sources & caveats
  • Source synthesis, not a similarity detector implementation or a measured compression result. [1]

JSON: /data/terms/resemblance-detection.json

Published by Hesela team · Updated

Report a correction · Editorial standards · Reuse guidance

Cite this page
Hesela team. Resemblance detection. Hesela. 2026-11-22. https://hesela.dev/terms/resemblance-detection/