The reduction of information in pictures is known as picture compression. The present work applies a psychophysical technique to evaluate the efficiency of compression algorithms. Two algorithms for stereo picture compression-one consistent with the suppression theory and the other with the fusion theory of depth perception-were compared and evaluated. Subjects estimated depth differences, and we found that these estimates under compressed conditions were as reliable as those under regular stereo conditions and significantly more reliable than estimates under mono conditions. We also found that picture compression results in a so-called depth recruitment phenomenon: the exponent of the psychophysical function was larger under compressed conditions than under the stereo condition. In contrast to performance in other tasks requiring depth perception, there was no difference in performance between the compression conditions.