To improve the penetration depth in photoacoustic microscopy while preserving high spatial resolution, we have proposed two-photon absorption-induced photoacoustic microscopy (TP-PAM). However, in tissue imaging, unwanted one-photon photoacoustic signals impair the image constructed from the two-photon photoacoustic signals, because the cross-section of two-photon absorption is smaller than that of one-photon absorption. To overcome this drawback, it is important to enhance (or extract) only the photoacoustic signals generated by two-photon absorption. In this study, to improve the detection selectivity and efficiency of two-photon photoacoustic signals, we investigated the dependence of TP-PAM signal intensity and image quality on the detection frequency range and excitation pulse duration in detail. The comparison among photoacoustic signals generated by optical pulses with various pulse durations (femtosecond to sub-nanosecond) enabled us to find that, the shorter the pulse duration is, the higher the generation efficiency of two-photon photoacoustic signals is. We also applied the confocal configuration between optical (excitation) and acoustic (detection) foci to TP-PAM. The optimization of the pulse duration, frequency filtering and confocal configuration improves the selectivity and efficiency of the TP-PAM signal. Such improvements can reduce the photon number required to obtain TP-PAM images and thus make the imaging speed faster and avoid tissue damage.