The present study recorded event-related potentials using rapid serial visual presentation paradigm to explore the time course of emotionally charged pictures. Participants completed a dual-target task as quickly and accurately as possible, in which they were asked to judge the gender of the person depicted (task 1) and the valence (positive, neutral, or negative) of the given picture (task 2). The results showed that the amplitudes of the P2 component were larger for emotional pictures than they were for neutral pictures, and this finding represents brain processes that distinguish emotional stimuli from non-emotional stimuli. Furthermore, positive, neutral, and negative pictures elicited late positive potentials with different amplitudes, implying that the differences between emotions are recognized. Additionally, the time course for emotional picture processing was consistent with the latter two stages of a three-stage model derived from studies on emotional facial expression processing and emotional adjective processing. The results of the present study indicate that in the three-stage model of emotion processing, the middle and late stages are more universal and stable, and thus occur at similar time points when using different stimuli (faces, words, or scenes).