FANN-based subsampling of color video sequences

In this paper, we present a subsampling method for color video sequences using feedforward artificial neural networks. In our method, subsampling of the high detail and smooth regions of the video chrominance frames is performed using pattern matching, by extracting different geometrical information for each region. Experimental results show that our method outperforms spatial subsampling obtained via lowpass filtering and decimation both objectively and subjectively. Moreover, our method significantly outperforms other FANN-based subsampling methods, especially at very low bit rates.