In Sparse Components of Images and Optimal Atomic Decompositions by Donoho, 1998, the author claims that
The human visual system is thought to do a tremendous job in achieving sparse representation of image data, taking $10^7$ bits/sec at the periphery of the visual pathway and winnowing it down to about 50 bits/sec deep inside.
Can somebody please point me to the source of that claim, as it is not mentioned in the paper, at least to my knowledge.