c++ - Perceptual hashing using Insight library -

for project working on make use of insight perceptual hash library. after reading interesting blog post (http://bertolami.com/index.php?engine=blog&content=posts&detail=perceptual-hashing) have tried make work myself, have little experience working image files in c++.

the code have far:

#include "stdafx.h" #include "vninsight.h" #include <string> #include <iostream>  int _tmain(int argc, _tchar* argv[]) {     cvimage ** a, ** b; // declare image objects     uint32 statusa = vncreateimage(vn_image_format_r8g8b8, 960, 960, a); // not initialise a?     uint32 statusb = vncreateimage(vn_image_format_r8g8b8, 960, 960, b);     uint8 * addressa = (*a)->querydata(); // documentation says have retrieve address function     vncompareimages(** a, ** b); // 'a not initialised error'     return 0; } 

the mess above conjured following documentation library, small enough can copy/paste here:

the primary interface insight compareimages function:

float32 vncompareimages( const cvimage & pa, const cvimage & pb ); function return value in interval of [0,1] indicates percentage of similarity between 2 input images.

cvimage objects simple wrappers around formatted memory buffer. create new image, call vncreateimage(), passing in dimensions , desired image format (currently vn_image_format_r8g8b8 supported insight).

once call completes, image object allocated , ready filled image data. can supply data copying address indicated cvimage::querydata(). don't forget call vndestroyimage() when you've finished image object.

my question is; how can simple image similarity check? moreover, cvimage object exactly? ("simple wrappers around formatted memory buffer" not entirely clear me)

i've used insight university project , had pretty decent results. code close, needs fixed in few areas:

cvimage *a, *b; // declare image object pointers vncreateimage(vn_image_format_r8g8b8, 960, 960, &a); // allocate space image vncreateimage(vn_image_format_r8g8b8, 960, 960, &b); // allocate space image b 

at point, , b each point allocated uninitialized images.

next need populate , b valid image data (otherwise comparing 2 black images). in project loaded data bitmap files , copied images.

once you've filled out images, call vncompareimages result.

float32 fpercentagematch = vncompareimages(*a, *b); 

don't forget deallocate images once you're finished them:

vndestroyimage(a); vndestroyimage(b);