Structural damage detection has been and will continue to be an important issue in many areas of engineering such as automotive, naval, aero and oil & gas industries. Many otherwise useful damage detection techniques are limited by a need for knowledge about healthy structure in order to discriminate damage. An existing algorithm, the Gapped-Smoothing method, which was developed for global non-destructive testing, has proved able to locate even small-sized damage in a beam-type structure without requiring any knowledge about the healthy structure. Modal curvature information is used in this technique to obtain a damage index (DI) which localizes the damage. As this information is derived from eigenvectors, it is imperative that the eigenvectors represent a noise-free form of mode shape. In practice, measured data contains noise that undermines the effectiveness of this technique. In this paper, reconstruction of the mode shapes are discussed along with the effect of sampling of mode shapes on the damage index. Then a noise suppression method is developed by considering two, three, four and five modes to improve the damage index. The results are analysed for both severe and small damage cases, on the steel beam with two slots.