This paper investigates four algorithms for dense matching of aerial images, they are (1) Photosynth, (2) Pix4UAV, (3) the SIFT-based approach and (4) tensor voting method. The afore-mentioned four algorithms are compared with respect to their dense matching performance inclusive of four issues: (1) accuracy, (2) reliability, (3) point density and (4) computation speed. This study adopts some aerial images covering a test field with high accuracy of check points. Error detection on their matching results will be done either by bundle block adjustment or relative orientation computation. The rate of successful matching will be assessed and analyzed. After blunder detection and deletion, the accuracy of the final matching results is evaluated by means of those ground check points. Some statistical figures are used to illustrate the quality and efficiency of these four dense matching algorithms.