A wavelet multi-resolution technique is applied to analysing the temperature field simultaneously obtained by a rake of 16 cold-wires in the turbulent near-wake of a slightly heated circular cylinder with a diameter of d = 12.7 mm in a range of x/d from 3 to 20, where x is the downstream distance from the cylinder axis. This technique enables us to decompose the fluctuating temperature field into a number of wavelet components based on different characteristic frequency bandwidths or scales, which are representative of the temperature fields of different scales. The turbulent mixing characteristics of various fluctuating scales are examined in terms of instantaneous temperature contours of each wavelet component. The flow structures and intermittent processing of various scales are visualized. The streamwise evolutions of temperature variance of various scales suggest that the intermediate-scale structures make larger contribution to the total temperature than the large- and small-scale structures. The wavelet auto-correlation function indicates that the large- and intermediate-scale structures display larger correlation and the wavelet component of higher frequency loses coherence quickly.