Stereo matching is a technique of finding the disparity map or correspondence points between two images acquired from different sensor positions; it is a core process in stereoscopy. Automatic stereo processing, which involves stereo matching, is an important process in many applications including vision-based obstacle avoidance for unmanned aerial vehicles (UAVs), extraction of weak targets in clutter, and automatic target detection. Due to its high computational complexity, stereo matching algorithms are one of the most heavily investigated topics in computer vision. Stereo image pairs captured under real conditions, in contrast to those captured under controlled conditions are expected to be different from each other in aspects such as scale, rotation, radiometric differences, and noise. These factors contribute to and enhance the level of difficulty of efficient and accurate stereo matching. In this paper we evaluate the effectiveness of cost functions based on Normalized Cross Correlation (NCC) and Zero mean Normalized Cross Correlation (ZNCC) on images containing speckle noise, differences in level of illumination, and both of these. This is achieved via experiments in which these cost functions are employed by a fast version of an existing modern algorithm, the graph-cut algorithm, to perform stereo matching on 24 image pairs. Stereo matching performance is evaluated in terms of execution time and the quality of the generated output measured in terms of two types of Root Mean Square (RMS) error of the disparity maps generated.