GetSimilarImageDeduplicationReportJSON

Image handling, Optimisation, Diagnostics

Description

Runs a read-only perceptual image deduplication analysis with the supplied thresholds and writes its work, accepted-error, and estimated payload-saving metrics as JSON

Syntax

Function TPDFlib.GetSimilarImageDeduplicationReportJSON(
  MaxHashDistance, MaxMeanError, MaxChannelError: Integer;
  Out Report: WideString): Integer;

Parameters

MaxHashDistance	Maximum 64-bit difference-hash Hamming distance from 0 to 7
MaxMeanError	Maximum mean decoded sample-channel error from 0 to 255
MaxChannelError	Maximum decoded error for one sampled channel from 0 to 255
Report	Receives the JSON document

Return values

Returns the number of similar image objects that the dry run would consolidate, or 0 when no pair is accepted or the request is invalid

JSON fields

thresholds	The requested `hashDistance`, `meanError`, and `channelError` limits
imagesScanned	Eligible indirect image streams examined
imagesDecoded	Images decoded into the fixed comparison sample
decodeFailures	Images skipped because their pixels could not be decoded
canonicalImages	Distinct accepted canonical groups
hashCandidates	Previously retained images sharing at least one difference-hash band
candidateComparisons	Pairs that passed semantic and hash-distance gates and reached sample confirmation
imagesDeduplicated	Images accepted as perceptual duplicates
referencesRepointed	References that a live operation would redirect
reclaimableBytes	Encoded payload bytes belonging to accepted duplicate streams
samplesCompared	Colour-channel sample values examined during confirmation
maxAcceptedHashDistance	Largest accepted hash distance
maxAcceptedMeanError	Largest accepted mean absolute channel error
maxAcceptedChannelError	Largest accepted individual channel error

Remarks

The method never rewrites references or removes objects

The report estimates encoded stream payload savings and excludes indirect-object, dictionary, cross-reference, and container overhead

Review the accepted maximum errors and resulting document visually before applying permissive thresholds