A new method for characterizing the speech perception performance of hearing-impaired listeners in both quiet and noisy environments is proposed. The concept of this method is that one's spech perception ability depends on the quality of one's perceived speech sound image. In this method, one's perceived image quality is related to subjectively a measured modulation transfer function (MTF). The procedure for estimating the sound image quality from measured MTF values obeys that of the RASTI method [Houtgast et al., Acustica 46, 60 (1980)], which is a physical method for predicting speech intelligibility in rooms. The subjective MTFs are obtained from the signal detection experiments using an adaptive 2IFC procedure with a modulated masker noise [Miyata et al., Acustica 73, 200 (1991)]. The proposed method was used to evaluate the monosyllable recognition abilities of six hearing-impaired subjects. The experimental results provide good support to the method for characterizing the speech perception ability of hearing-impaired listeners.