Using Perceptron Neural Network and Genetic Algorithm for Image Compression and Decompression

Abstract

This paper introduces an idea for image compression by using Genetic Algorithm and Arithmetic Coding. First stage is by using perceptron neural network to compress each three-pixels into single value, this value is called Compression value. In this paper the neural network did not need weights for training at compression part, the weights used in this paper are of one dimensional array containing floating point values. The total of weights values equal one. In the decompression part we use Genetic Algorithm for return the pixels, by using Crossover operation and the Fitness Value. The fitness value represented the error between Compression value and Desired output for each generated string by GA to get approximately original three-pixels. The second stage is Arithmetic coding algorithm uses to convert vector of compression values into a single floating point number. Our approach tested with color image, also in this paper the performance of the algorithm is computed and compared wit PNG and TIFF algorithm.