The absorption band center of visible and near infrared reflectance spectra is a key spectral parameter for lunar mineralogical studies, especially for the mafic minerals (olivine and pyroxene) of mare basalts, which have two obvious absorption bands at 1000 nm (Band I) and 2000 nm (Band II). Removal of the continuum from spectra, which was developed by Clark and Roush and used to isolate the particular absorption feature, is necessary to estimate this parameter. The Moon Mineralogy Mapper (M-3) data are widely used for lunar mineral identification. However, M-3 data show a residual thermal effect, which interferes with the continuum removal, and systematic differences exist among optical data taken during different optical periods. This study investigated a suitable continuum removal method and compared the difference between two sets of M-3 data taken during different optical periods, Optical Period 1B (OP1B) and Optical Period 2A (OP2A). Two programs for continuum removal are reported in this paper. Generally, a program respectively constructs two straight lines across Band I and Band II to remove the continuum, which is recommended for locating band centers, because it can find the same Band I center with different right endpoints. The optimal right endpoint for continuum removal is mainly dominated by two optical period data at approximately 2480 and 2560 nm for OP1B and OP2A data, respectively. The band center values derived from OP1B data are smaller than those derived from OP2A data in Band I but larger in Band II, especially for the spectra using longer right endpoints (>2600 nm). This may be due to the spectral slopes of OP1B data being steeper than those of OP2A data in Band I but gentler in Band II. These results were applied to Mare Humorum and Mare Nubium, and the measurements were found to mainly vary from intermediate-to high-Ca pyroxene.