A fully interconnected neural network approach and its applications in image processing

In previous works, back-propagated neural network (BPNN) had been applied successfully in the spectral estimation and in the spatial resolution improvement of remotely sensed low resolution images using data fusion techniques [1-4]. Besides, other types of learning algorithms have been proved their validity in image denoisification, enhancement and classification [5]. However, the time required in the learning stage has been long, particularly in the applications of BPNN. In the present paper, a fully interconnected neural network model is developed. With this model the global minimum error is reached considerably faster than with any other method without regarding the initial settings of the network parameters.