Image catalog - fuzzy match with percent score
This compares the query file against the catalog using a block-level content fingerprint
(16 chunks, each chunk hashed separately) and reports what percentage of chunks match. It catches
near-duplicates at the byte level (same image re-saved with different metadata, a few edited
bytes) - it is not visual similarity/computer-vision matching: a resized, cropped, or
recompressed copy of a catalog image will score low or 0%, because compression changes every byte,
not just a few chunks. Real visual-similarity matching needs image-decode + perceptual-hash or DNN
feature-extraction infrastructure, which is separate, not-yet-built work (see
WEBSUPPORT_ICPP_PLAN.md Part 10's ivec_* vector search engine for where that would plug in).
Catalog
| id | filename | size | hash |
|---|
", icpp_sql_get($q,"id"), " | ", icpp_sql_get($q,"filename"),
" | ", icpp_sql_get($q,"filesize"), " | ", icpp_sql_get($q,"filehash"), " | ");
}
icpp_sql_free($q);
?>
Query result
Query file: ", icpp_getparam("query_path"), "");
iprint("| catalog image | percent match |
|---|
");
$m = icpp_sql_query($db, "select * from image_catalog;");
while(icpp_sql_next($m))
{
$pct = icpp_fingerprint_similarity(icpp_file_fingerprint(icpp_getparam("query_path")), icpp_sql_get($m,"fingerprint"));
iprint("| ", icpp_sql_get($m,"filename"), " | ", $pct, "% |
");
}
iprint("
");
icpp_sql_free($m);
icpp_sql_disconnect($db);
?>