TImageEnMView.FindSimilar

Declaration

function FindSimilar(Index: Integer; Threshold: Double = 0.75; Algorithm: TIEVisionImageHashAlgorithm = ievHashPHash): TIEArrayOfInteger; overload;
function FindSimilar(Index: Integer; var Similarities: TIEArrayOfDouble; Threshold: Double = 0.75; Algorithm: TIEVisionImageHashAlgorithm = ievHashPHash): TIEArrayOfInteger; overload;

Description

Returns the indexes of all images that are the same as, or similar to, the image at Index (requires IEVision).
Images are compared using a perceptual hash, so resized, recompressed and slightly edited copies are found.
The result is sorted with the most similar image first. Index itself is not included.

Parameter	Description
Index	The image to compare with all others
Similarities	Returns the similarity (0 to 1) of each returned image, in the same order
Threshold	Minimum similarity (0 to 1). 0.75 finds near-duplicates. Use a lower value to find more loosely similar images. See similarity
Algorithm	The perceptual hash algorithm. See TIEVisionImageHashAlgorithm

Works with any StoreType: images loaded on demand are loaded to be hashed, and thumbnails are hashed as they are (which is faster and gives the same result).

Note: Every image is hashed on each call, so for repeated searches of large collections it is faster to store the hashes (GetPerceptualHash) and compare them with IEVisionCompareHashes.

Demo

Demos\ImageEditing\EveryMethod\EveryMethod.dpr

Example

// Check all images that are similar to the selected one
var
  idx: TIEArrayOfInteger;
  i: Integer;
begin
  idx := ImageEnMView1.FindSimilar( ImageEnMView1.SelectedImage );
  for i := 0 to Length( idx ) - 1 do
    ImageEnMView1.Checked[ idx[ i ]] := True;
end;

TImageEnMView.FindSimilar