Halftoning Technique Using Genetic Algorithm Naoki Kobayashi and Hideo Saito 1994 IEEE.

Halftoning Technique Using Genetic Algorithm

Naoki Kobayashi and Hideo Saito1994 IEEE

The Flow of GA

Halftoning Technique using Genetic Algorithm

There are two proposed methods for halftoning techniques using GA.

Method I-Initial Population

The initial population of n strings such as the sample is produced randomly and independently of the gray-tone block.

Method I-fitness value

bg(i,j) is a convoluted b(i,j) by the gaussian filter. gs(I,j) is the convoluted g(I,j) by the smoothing filter.

blockji

gm jibjigs

E),(

|),(),(|1

blockji

sc jibjigM

jigs

E),(

|),(),(2

),(|1

ccmmt EwEwE

tf EEF

Method I-fitness value(cont.)

Method I-Reproduction

The population size is n and the fitness value of the ith string is F(i), a selection’s probability P(i) of ith string

n

j

jF

iFiP

1

)(

)()(

Method I-Crossover

One of the two crossover methods is selected randomly.

Method I-Mutation

The string is selected randomly and one pixel in the string is inverted. The black changes to the white or the white changes to the black.

Method II-Initial Population

The Nb black pixels

M

jig

N blockjib

),(

),(

Method II-Fitness Value

The calculation is the same as it in the method I.

Method II-Reproduction

The calculation is the same as it in the method I.

Method II-Crossover

The number of black pixels between tow strings are changed when the pixels in the local region are exchanged, some pixels are randomly selected and then exchanged between two strings for making the number of black pixels even.

Method II-Mutation

The string is selected randomly, two pixels in the string is selected randomly and the two pixels are exchanged.

Conclusion

Using the proposed method, visually pleasing halftone images were obtained.

However, the computation time was long in this method.