An artificial neural network is proposed which is able to compress an image by computing a nonlinear nonorthogonal transform and its inverse. The network is trained with small blocks extracted from an image; after the learning phase, it proves to give satisfactory performances both for the learned and for different unlearned images.
No takes yet. Share an insight, caveat, or question.
Sicuranza et al. (1990) studied this question.
Synapse has enriched 5 related papers. Consider them for comparative context: