OBJECTS IMAGES ALIGNMENT WITH THE USE OF GENETIC AND GRADIENT ALGORITHMS
The given paper presents a hybrid method which is a combination of genetic and gradient algorithms used for the comparison of digital images of an object. Aligning the images, their basic transformations are taken into account, namely shift and scale in a width and height, angle, changes in intensity and contrast. The software for image alignment of objects has been created using Delphi environment. The program utilizes modified genetic algorithms where the chromosomes are the vectors of real numbers. The methods of roulette, rank and tournament selection are used for chromosome selection. After the use of the genetic algorithm the object images were compared by the method of coordinate descent that provides an accuracy improvement of image alignment. The efficiency of different methods of chromosome selection in the genetic algorithm for images alignment is researched. The size of chromosome population as well as other parameters of genetic algorithm have been optimized.