When a narrow-band source propagates in a shallow ocean environment, the ocean can be characterized by a waveguide resulting in a normal-mode solution. When the source is broadband, its spectrum can be decomposed into a set of distinct narrow-band lines, each supporting a distinct set of normal modes differing from frequency to frequency by both the number of modes and their respective wave numbers. Couple this propagation problem to a set of noisy hydrophone measurements obtained from a vertical array of sensors spanning the water column and a formidable problem in estimating the broadband acoustic field results. The field itself is essentially a space--frequency and once this space--frequency field can be estimated from the noisy measurements, then various postprocessors can be applied to estimate, detect, and localize the underlying source. Thus it is important, in the broadband case, to extract and enhance the broadband acoustic-field image. In this paper the development of a model-based image enhancement technique is discussed which employs the broadband normal-mode solutions using an optimal estimation scheme to provide the required enhancement. A by-product of the outputs of the processor, which is a space-varying Kalman filter, is the ability to not only enhance the image, but also to extract and enhance the corresponding broadband modal functions (one for each frequency) as well.