J. FZuid Mech. zyxwvutsrqp (1982), vol. 119, zyxwvuts pp. 323-345 zyxwvu Printed in Great Britain 323 Computer analysis of a high-speed film of the plane turbulent mixing layer By MIGUEL A. HERNANT AND JAVIER JIMENEZt School of Aeronautics, Universidad Politecnica de Madrid, Spain UAM-IBM Scientific Center, P. Castellana zyxw 4, Madrid, Spain (Received 15 April 1981 and in revised form 18 November 1981) To evaluate the usefulness of digital image analysis in extracting quantitative infor- mation from flow pictures we have studied a zyxw 16 mm cine film of a turbulent mixing layer. A sequence of 373 frames is digitized and analysed to isolate and measure the concentration eddies that constitute the large structure and to follow their individual evolution in time. As a result, statistics are given on the life history of the eddies, the structure of the amalgamation process and the amount of entrainment, as measured by area change, due to amalgamation as compared to the total. It is found that most of the entrainment occurs during the normal life of eddies and not during pairing. Mixing intermittency is computed from the observed shape of the eddies and seen to compare well with previous measurements. The significance of these results in model- ling the mixing layer is discussed briefly and some comments are given on the general usefulness of the techniques presented. 1. Introduction Visual information has always been important in fluid mechanics. Nowhere has this been clearer than in the study of the large-scale structures in turbulent shear flows. Although it had been known for some time that large eddies were important in those flows and that their behaviour was not strictly disordered (Townsend 1956), the full realization of their quasi-deterministic nature and of their importance in ordering the flow had to wait for the beautiful shadowgraphs contained in Brown zyx & Roshko (1974). Unfortunately it has been difficult up to now to extract more than qualitative information from flow-visualization pictures. The reasons are several : firstly the photographic processing of pictures is slow and cumbersome, secondly it is difficult in many cases to obtain accurate measurements of optical density and dimensional data from pictures, and finally the amount of information, relevant or irrelevant, contained in a single image, and even more so in a cine film, is so great as to tax the processing ability of a human observer. Automatic digital image processing has been developed in the last two decades to cope with simiIar problems in areas like remote sensing and medical technology, and t Present address: Department of Aeronautics, California Institute of Technology, Pasadena, U.S.A. $ Present address: Department of Applied Mathematics, California Institute of Technology, Pasadena, U.S.A.