Rembrandt

I rediscovered the art of the dutch artist Rembrandt van Rijn (1606-1669) while visiting the Hermitage museum in St. Petersburg. Here is one of his many paintings that are on display there.


Abraham and Isaac 1634.

The picture above of course does not do justice to the painting itself. What is interesting about Rembrandt's art is that it can convey an enormous amount of information about what is seen in the picture with very little detail. When peeking very closely at one of his paintings (especially the later ones) one is surprised to find very coarse blobs of painting. Rembrandt supposedly said about his art: "Do not get too close to my paintings, the smell of the paint will sicken you."

It is very interesting that more detailed paintings do not appear as realistic as Rembrandt's. Some of Rembrandt's portraits (most notably his autoportraits) seem to capture the subject more intensely than any photograph. Somehow one has the feeling of staring at a real human being.

This may seem paradoxical at first. One may think that the more information we put into the painting (i.e., visual detail) the higher the realism. This reasoning, however, underestimates the active role of our vision. The less detail, the more our visual system has to "fill in the missing information". This is like trying to spot a familiar objects in a hazy environment. Since we know we are looking at a portrait we somehow expect to see a face. Rembrandt probably knew this, and he has found the right tradeoff between too much and too little detail. There should be enough detail such that we "recognize" a face, but not enough such that we see it as a "surrogate copy" of the face.

This should be of importance to anyone interested in creating realistic pictures. Why increase the resolution of a display (high resolution TV) when a coarser resolution actually does a better job ?

Instead of trying to throw more computing power and memory at computer graphics simulations, maybe we should study the effects of "textures" on our visual system more closely. Of course the latter cannot be cast into a physical equation.