Neural networks for image modeling by two-dimensional random fields with application to image compression for target acquisition
The problem of target acquisition is considered to be a very involved process and is a serious challenge for the researchers. For several applications of target acquisition, it is worthwhile to compare the compressed and uncompressed images and the perceptual difference between the two images is also significant. A new neural network technique of image modeling by 2D random fields formulated in the form of an autoregressive moving average process driven by input white Gaussian noise with known statistics is presented. The proposed technique consists of two stages: (1) estimating the parameters of the model and (2) regeneration of the image with the knowledge of the model, its parameters, initial conditions, and white noise. The problem of estimating the model parameters is formulated as an optimization problem solved by a single-layer neural network. Once the model parameters have been estimated as the adaptive weights of the network, the second stage reconstructs the picture from the model. This stage consists of recursively constructing the image using the initial conditions of the original image, the parameters of the model, and white Gaussian noise. Due to the adaptive nature and the computational capability of the neural network, a high-quality image is obtained with this approach. The proposed algorithm reduces the computational complexity and is recommended for the on-line image compression required in target-acquisition-type applications. As the image is constructed using fewer pixel values of the given image in the form of initial conditions, and a few parameters of the model, very effective image compression is achieved. Several computer simulation examples are included to illustrate the effectiveness of the proposed technique.