Colour-Science Similar Image Finding (SIF) technology

A slightly different approach compared to the automatic images classification (AIC) is to find similar images once you have a sample image.

The idea is that you browse through your images and find one image you like. Now you want to see if there are similar images to this image. The SIF software now sorts your images so that the most similar images will be first.

SIF uses image classes but also techniques which rate the general similarity of images. This similarity is a combination of color and density histogram correlation and the correlation of important details.

This project is still under development but we have working prototypes to show how similar images can be found.

Screenshot of the Colour-Science similar image finder software

Here is a screenshot of the output of a selection of sorted images. With the test software the images now have a number as filename. The number is the difference to the sample image. We find the sample image with a difference of 0 as the first image. This is because the sample image was also part of the images

Now here is a screenshot of the most similar images:

here inbetween are lots of other images ………………..

these are the images which the biggest difference to the sample image