Recently, the application of a scale-space filter for speckle noise reduction in electronic speckle pattern interferometry correlation fringes was reported. The filter, derived from information theory and statistical mechanics considerations, is an iterative, adaptive, and clustering algorithm that performs edge-preserving smoothing. To apply the filter, the kernel size and the number of iterations must be previously set. We show that both parameters can be chosen depending on image features, e.g., presence or absence of sharp edges due to shadows, holes, or physical limits of the test object. When high-contrast edges are present, it is also shown that better noise reduction and edge preservation are obtained when the convergence of the solution of the filter is truncated. Both computer-simulated and experimental correlation fringes are used to test the scale-space filter.