A neural network approach that learns to represent images as continuous functions rather than discrete pixels.