This paper concerns the possibilities for side scan sonar to determine the bathymetry. High definition sea-bottom imaging is the primary task for new side scan sonars; nevertheless, these systems are also able to estimate the relief with a definition equal to the pixel resolution of conventional imaging sonar, through an interferometric multisensors approach. The main drawback concerns both the errors and accuracy of the computed numerical elevation model. Indeed, interferometric methods use a phase difference to determine a time delay between two sensors observing the same sea-bottom location. The phase difference belongs to a finite interval ] -pi ,+pi], while the time delay between two sensors does not: this mapping induces a 2pi biased phase. Very often, sonar are well-designed to allow removing this bias through vernier techniques. However, some difficulties come from interferometric noise, which generates errors on the 2pi bias estimation derived from the vernier. A traditional way to reduce noise impact on the interferometric signal is to average received data. Obviously, such a method does not preserve the resolution of the bathymetric estimation. Thus, this paper presents an attempt to improve both the accuracy and resolution of the interferometric signal relying on a wavelet based method of image speckle reduction. Traditionally, speckle noise is processed as the logarithm of the signal absolute value. However for this application, the proposed interferometric speckle reduction is achieved directly on the interferometric signal by integrating information, with the guidance of the despeckled image. On the presented results, phase despecklisation improves considerably the quality of the interferometric signal in terms of signal to noise ratio, without an important degradation of resolution