CWT is based on correlation of wavelets and given signal over a fixed time. For example, if signal time length is 1&ie and wavelet length is 1sec.. Then CWT provide correlation indices after every 1sec: resulting in 60 indices. Inference
Yes, I agree that the above recommendation for the article written by Torrence and Compo (1998) is a helpful article. However, to interpret the scalogram (the 2-D plot of time vs. frequency where the contour of power, the lower of the two plots you showed) one needs to know the shape of the wavelet used to generate the coefficients contoured in the scalogram. For instance, if one used the Morlet wavelet (sinuoidal), then the red zones in the plot would indicate the time and frequency that have sinusoidal components. However, if a different wavelet was used, then the interpretation would be different.