We have developed an algorithm to filter the noise in the spectral intensity of ultrashort laser pulses. The filtering procedure consists of smoothing the noise by using the Savitzky-Golay filter, removing the offset, and using a super-Gaussian window to truncate the frequencies of the spectrum. We have modeled bandwidth-limited ultrashort pulses with Gaussian modulated frequencies to show the estimation of the carrier wavelength, reconstruction of the intensity pulse profile, and pulse duration after applying the algorithm. Theoretical results are presented for pulse durations between 5 fs and 100 fs with a carrier wavelength of 825 nm and three different amounts of signal-to-noise ratio (SNR): 30 dB, 20 dB, and 15 dB, normally found in experiments. The algorithm is also applied to an experimental spectral intensity from a homemade Ti:sapphire laser that produces pulses of about 20 fs at 825 nm at 100 MHz. We will show that using only a low-pass Fourier filter and removing offset is not enough to recover the spectral intensity when a large SNR is present, which may be the case when the ultrashort laser beam has been manipulated to compensate for the group velocity dispersion of an external optical system. In cases like this, the use of the Savitzky-Golay filter prior to the super-Gaussian filter improves the recovery of the carrier wavelength and the spectral intensity. We will also show that the algorithm presented in this paper is suitable for experimental analysis and requires limited user intervention.