Disminución de ruido en imágenes satelitales en el sistema DVB-S2

Abstract

In this article, the simulation of the DVB S2 standard (digital video broadcasting by satellite version 2) implemented in the simulation tools Simulink and Matlab for the elimination of noise in satellite images is carried out, using a convolutional neural network. QPSK (Quadrature Phase shift keying) modulation was used within the block diagram of the DVB-S2 standard, to add noise to the images, this value being modified in the channel AWGN of - 2.75 [dB]. For noise level checking the output images of the DVB-S2 system are compared with and without CNN (Convolutional neural network) filter. Convolutional neural networks have superior performance in many visual tasks, such as classification and object detection. Through CNN training an improvement of 6.49% was obtained, this represents that with the use of CNN noise in images is substantially eliminated.