Spectral calibration is an essential step in the calibration process of multispectral camera for Mars exploration. The precision of spectral calibration has a direct impact on the accuracy of Mars multispectral data inversion. The most common method for determining the central wavelength and spectral bandwidth of a multispectral camera is to fit the spectral response using a Gaussian function. The peak value fitted by Gauss is defined as the central wavelength of the camera, and the full width of the spectral band at half maximum is defined as the spectral bandwidth. But the spectral response of the Mars exploration multispectral camera in the near-infrared waveband deviates from the Gaussian distribution. In this case, the relative output deviation and the relative radiation error are large when data inversion is performed using the central wavelength and the spectral bandwidth determined by Gaussian fitting. In this paper, the spectral response of a multispectral camera is taken as a weight function, and a generalized weighted method is proposed to redefine the central wavelength and spectral bandwidth of the multispectral camera. The central wavelength is defined as the weighted average of wavelengths against the spectral response weighted factor, while the spectral bandwidth is defined as the generalized w bandwidth when the area within the bandwidth is w% of the total area. The simulation results showed that the relative output deviation was reduced by an average of 0.1% and a maximum of 1.3%, while the relative radiation error by an average of 0.14% and by a maximum of 1.83%, when compared with the Gaussian definition. The experiment results of the Mars exploration multispectral camera showed that there was an average reduction of 0.06% in relative output deviation and 0.05% in relative radiation error. The generalized weighted method provides another way for the definition of central wavelength and spectral bandwidth, and improvs the precision of Mars multispectral data inversion.