From Pattern Recognition to Image Understanding (Language, Segmentation, Concepts) 1
Shelia Guberman, Vadim V. Maximov, ALEXANDER PASHINTSEV · 2010
To resolve the image search problem, an algorithm has to understand images, i.e., be able to describe their content using an adequate language. With that approach the comparison of images becomes comparison of descriptions. To accomplish the above goal the software was developed,, which at the first step segmented the image. The segmentation is based on Dual Clustering procedure, which generates a limited number of segmentations and chooses the best according criteria. At the second step segments are recognized as notions sky, vegetation, water, ground, mountains, buildings and some more.