Neural Network based Soft Computing Technique for Image Compression and Decompression

Image compression is the technique used to minimize memory space and decrease bandwidth (reduce high data rate) for transmission without deteriorating image quality. The various methods and standards for image and video like JPEG, Wavelet, M-JPEG, H.26x etc. have been proposed by researchers. Even though, Increase in mass-storage density, speed of processor, and digital communication system performance, demand for data storage capacity and data-transmission bandwidth continues to outrage the capabilities of available technologies. Apart from the above mentioned existing technology on image & video compression, a method of image compression using soft computing have been proposed. Two layer Feed forward neural network will be considered and will be trained off-line using Levenberg Marquardt algorithm. The weights of trained network of hidden and output layers are used for image compression and decompression respectively for any test image. MATLAB will be used as software tool to carry out for training neural network, image compression and image decompression. The performance parameters like compression efficiency, complexity of algorithm and quality, for image compression & decompressions will be analyzed.