David Marr (neuroscientist)

David Marr ( born January 19, 1945 in Essex, † November 17, 1980 in Cambridge (Massachusetts ) ) was an English psychologist, computer scientists and mathematicians. He is considered one of the founders of neuroscience and computer science developed a model of vision as an information processing of the brain that functions according to the principles of the electronic data processing on a computer. The Marr Prize was named after him and is every 2 years awarded on the ICCV conference.

Life

Educated at Rugby School Marr studied mathematics and computer science neuroscience at Cambridge University. Later he worked at the Massachusetts Institute of Technology. Terminally ill with leukemia, he wrote with great effort at the end of his life, his major work, Vision, published only two years after his death with the help of many colleagues.

Work

Marr linked results of psychology, artificial intelligence and neurophysiology to a new model of the functioning of vision. For him, seeing is an information processing of the brain that does not operate with images whose light is already stopped on the two -dimensional retina and converted by it into neural activity. These electrical activity patterns represent the outside world and are combined only in the brain to what we call vision. To understand this process, one must clarify by Marr questions at three levels:

  • The computational plane (computational level ): What purpose does the seeing? What can you afford?
  • The algorithmic level ( algorithmic level ): How can the goal of the computational level can be achieved, ie as may be shown accurately input and output data, and what is the algorithm that transforms the input data in the output data?
  • The technical level ( implementational level ): How is the algorithm implemented physiologically or physically neuronal activity?

These distinctions Marrs also gained importance in other disciplines, in the design and analysis of artificial neural networks in the sequence.

Seeing as the processing of two-dimensional data of the retina ( retina) to a three-dimensional description of the world is according to Marr, a process that takes place in four stages:

  • It is based on the retinal image, a projection of the outside world, which includes brightness or color information in the form of an intensity image;
  • The first draft ( " sketch " sketch ) are in the recognized by analyzing the changes in brightness edges, contiguous areas and surface textures like a line drawing;
  • The 2 ½ -D design that is created by combining the two retinal information and the spatial orientation and rough depth of surfaces determined, so that a first rough picture of the three-dimensional world emerges.
  • The 3 -D model, which gives a precise information about the depth of the lateral disparity of the two retinal images so that the perception in spite of eye and body movements remains stable.

The Marr wavelet

Although Marr's model of vision is not without controversy, as it provided as a byproduct an important contribution to until the end of the 1980s, evolving theory of wavelets. In particular, for the detection of edges, the zeros of the second derivative of the intensity function are crucial, according to Marr, so their turning points, see figure. Since the retina provides a two-dimensional image, this approach leads to the second mathematical derivative of the two dimensional Gaussian distribution as a filter function,

Today we call this function often Marr wavelet, it is multiplied by (-1) and so-called because of its shape Mexican Hat:

Cambridge (Massachusetts) England Neuroinformatics International Conference on Computer Vision Integrated Authority File
221921
de