The problem of the evaluation of optical images is a very old one. However, it has never been solved in a satisfactory way, simply because it has never been conveniently stated. Before attempting to evaluate an optical image, one should answer the question: What do I want? One and the same image can be very good for a given purpose and very bad for another purpose.
The right way of tackling these problems has been found only in recent times. Information theory enables one at least to state the problem in a clear and unambiguous way. The work carried out by researchers in this field (Blanc-Lapierre, King and Emslie, Linfoot, Fellgett, H. H. Hopkins, and several others) already authorizes the hope that the question will very soon have a quantitative answer in each particular case. The answer will be represented by a number of bits, measuring the capacity of the channel.
In communication theory the capacity of a given channel depends on the statistics of the source of information. Researchers in optics will necessarily be confronted with the difficult problem of specifying the statistics of the object. There is a feeling that present optical instruments, which are designed to meet a much more general statistics than that of actual objects, show a tremendous redundancy. In other words, a large portion of their capacity is wasted.
The problem of reducing the redundancy is closely related to that of encoding the information. That the information contained in an optical image can be encoded is proved by the existence of the technique of drawing and of other pictorial techniques. Drawing was one of the first codes discovered by mankind to spare information. The number of bits contained in a drawing is incredibly less than the number of bits contained in an image of the same object formed by a conventional optical instrument giving the same sharpness of the contours. However, for most purposes the drawing is sufficient because the observer combines the information carried by the drawing with the prior information which he has about the statistics of the object.
Perhaps the future of optical research will be to develop suitable methods of encoding optical information. A combination of optical and television techniques seems to be very promising and will possibly help to reduce the redundancy.
© 1957 Optical Society of AmericaPDF Article