In imaging, much attention is paid to increasing the resolution capabilities of a system. Increasing resolution allows for high-accuracy source location and the ability to discriminate between two closely-spaced objects. In conventional narrowband techniques, resolution is fundamentally limited by the size of the aperture. For apertures consisting of individual elements, direction-of-arrival techniques allow for high-resolution images of point sources. The main limiting factor on conventional high-resolution imaging is the number of elements in the aperture. For both passive and active imaging, to resolve K point sources/targets, there must be at least K+1 elements receiving radiation. In active imaging, when these targets reflect coherently - the more difficult case in imaging - an additional constraint is that at least K of the elements must also be transmitting radiation to illuminate the targets. For small arrays consisting of only a few elements, this constraint can be problematic. In this dissertation, we focus on improving resolution by using multiple frequencies in both passive and active imaging, especially for small arrays. Using multiple frequencies increases the size of the coarray, which is the true limiting factor for resolution of an imaging system when virtual arrays are considered. For passive imaging, we show that the number of sources that can be resolved is limited only by the bandwidth available for certain types of sources. In active imaging, we develop a frequency-averaging method that permits resolution of K coherent point targets with fewer than K transmitting and receiving elements. These methods are investigated primarily for linear arrays, but planar arrays are also briefly examined. Another resolution improvement method researched in this work is a retransmission scheme for active imaging using classical beamforming techniques. In this method, the coarray is extended not by using multiple frequencies, but by retransmitting the received data back into the scene as a second transmission and processing the returns. It is known that when this method is used to image multiple targets, the resulting image is contaminated by crossterms. We investigate methods to reduce the crossterms.