Object recognition, detection and segmentation tasks using RGB-D data are frequently used in the field of machine vision and robotics. In this work, a method for object segmentation on RGB-D data is proposed. In the first step, filtering steps are applied to obtain a better representation of the depth data. Surface normals are calculated for all points in the depth data. In the next step, a modified version of the region-growing algorithm is used, in which surface normals are compared. Thus, surfaces of objects are obtained. In the last step, with the help of the features extracted from the surfaces, objects are segmented by clustering surfaces with X-means. The proposed method has been tested with data sets containing complex scenes and is comparable to the related studies in the literature.