Method and device for measuring similarity between images
View Patent ↗A method is disclosed for measuring similarity between images, comprising the performance of the following steps for each of the images: segmentation of the image into segments, classification of the segments as a function of their orientation to give classes, calculation of a histogram of the number of segments as a function of class, calculation of a histogram of the number of pixels belonging to the segments of one and the same class as a function of class, and comparing, for each image, the histograms calculated above to their respective histograms calculated for each of the rest of the images to give a measurement of similarity.
1. Method of measuring similarity between images, comprising:
performing, for each of the images, the following steps:
(a) segmentation of the image into segments,
(b) classification of the segments as a function of their orientation to give classes,
(c) calculation of a histogram of the number of segments as a function of class,
(d) calculation of a histogram of the number of pixels belonging to the segments of one and the same class as a function of class,
comparing, for each image, the histograms calculated above to their respective histograms calculated for each of the rest of the images to give a measurement of similarity.
2. Method according to claim 1 , also calculating a third histogram corresponding to the distribution of the segments about the centre of gravity of each class.
3. Method according to claim 2 , wherein, to calculate the third histogram, it performs a calculation of the standard deviation of the distances from the middles of the segments of a class to the centre of gravity of the class considered.
4. Method according to claim 1 , characterized in that wherein the comparison of the histograms consists of a subtraction of the ordinates class by class and of a sum, over the set of classes, of the subtraction result obtained for each class.
5. Method according to claim 1 , wherein the histograms are coded according to the MPEG-7 standard.
6. Method of clustering images of a database, wherein the clustering is performed as a function of the measurements of similarity according to the method of claim 1 to give clusters of images.
7. Method of creating video summaries, implementing a selection of at least one of the images of at least one cluster calculated according to the method of claim 6 .
8. Method of video indexation, implementing a selection of at least one of the images of at least one cluster calculated according to the method of claim 6 , as indexation image.
9. Device for measuring similarity between images, comprising a circuit capable of
receiving digital data defining the images
performing, for each of the images, the following steps:
(a) segmentation of the image into segments,
(b) classification of the segments as a function of their orientation to give classes,
(c) calculation of a histogram of the number of segments as a function of class,
(d) calculation of a histogram of the number of pixels belonging to the segments of one and the same class as a function of class,
comparing, for each image, the histograms calculated above to their respective histograms calculated for each of the rest of the images to give a measurement of similarity.