The surface topography and morphology of high grade office and print papers is of immense importance for high printing performance and a constant product quality. As a result, reliable analytical methods are necessary to assess the influence of production parameters. In addition, it is of high interest to analyze the influence of paper morphology on the printability of the products. Several surface sensitive methods have been used in the past to investigate and illustrate high-resolution topography of paper. One of these methods is confocal laser scanning microscopy (CLSM), which provides nanometre depth precision. Even though CLSM is a very powerful technique that is frequently used in life sciences, sample preparation and data analysis are far from being trivial. In order to define the surface, images are stacked one over the other. At each XY position, the column of intensity values is inspected and the Z coordinate of the surface is defined by the voxel with the highest intensity. Although this surface definition gives relatively good surfaces, it still contains noise. Additional filtering is therefore required in order to remove small undesirable peeks and holes, which appear as a consequence of thresholding error. However, when smoothing the data, we need to be careful and remove only the noise without damaging the actual surface. We achieve this by estimating heights and widths of peeks and removing only those peaks, which are too small to represent the fibres. For this purpose, we filter the surface with a series of different filters, ranging from the smallest one to the largest one. The filter that removes the highest peek at a particular pixel is obviously the most important one and it defines the height and the width of the peak. Since fibres are of round shape, they should produce lower peaks at highest scales, while noise is characterized by sharp high peaks. Based on this, we can efficiently delete these values and replace them with interpolated values. For interpolation, we use an inverse distance weighting method, as it does not introduce any additional variations. Denoised surface is then ready for evaluation of roughness. However, the paper is not completely flat when it is being scanned and thus, the output surface is concave. In order to minimize this effect, the average height within the neighbourhood of each pixel is estimated and subtracted from the height of the pixel. The obtained difference is then used to calculate the root mean square value that is essentially the estimation of paper roughness.