The paper proposes an application of the median form ambiguity function in direct sequence spread spectrum modulated signals denoising. The observed signals are multicomponent and consisted of short duration sinusoidal components, appearing on different frequencies. The analysis of such multicomponent signals in the time-frequency plane could be disturbed by the unwanted cross-terms. It is shown that the filtering based on the median ambiguity function can completely eliminate the cross-terms. Moreover, beside the cross-terms the observed signal can be disturbed by different types of noise. Impulse noise is considered, due to the fact that this type of noise is common in communications. The signal terms are located around the origin in the ambiguity plane, being symmetric around the y-axis. Since the noise/cross-terms are dislocated from the origin, the filtering does not affect the signal terms providing satisfactory results. After denoising and cross-terms removal, the signal parameters (hop bandwidth and hop time duration) are estimated. The theory is proved by the experimental results.