Abstarct: Images formed out of the interpretation of ground gravity data are used to discover mineral resources, cavities, hydrocarbon reserves, etc. In these images, it is crucial to detect the boundaries and edges of gravity data which are the result of gravity difference at those points. Presence of noise and aliasing makes it difficult to interpret these images; therefore, the effect of disturbing factors such as sources of noise and the impacts of depth, source slope, and the overlap between sources which impedes correct interpretation must be undermined or modified by using different filters. Different filters, of which no single one can have all the characteristics of others, are used in edge detection of potential field anomalies. This study develops the generalized derivative operator filter to detect the boundary of anomalous sources such as empty underground spaces which are caused by gravity difference at those points. The results of application of this filter to the gravity data are compared to the results obtained using other filters such as analytical signals, total horizontal derivative, theta map, etc. For this purpose, first the required codes and functions are obtained using MATLAB software and, afterwards, the filters are applied to synthetic models so that the capabilities of each of the filters in detecting the boundaries of anomalous sources are determined. By applying the filters to the gravity data obtained by the synthetic models, it is found that the generalized derivative operator filter yields good results in comparison to other filters. Considering that filters have different efficiencies when dealing with actual gravity data, the study carries out gravimetric operations on Dehmolla Educational Mine Tunnel, Shahrood, Iran. It should be noted that in this study empty spaces are the dimensions and boundary of Dehmolla Educational Mine Tunnel which can be studied as an actual underground cylindrical model. After correcting the data obtained, by applying different filters to the gravity data of Dehmolla Educational Mine Tunnel, it is found that the generalized derivative operator filter generalizes some of the effects of analytical signal filters, total horizontal derivative, theta map, etc. The filter’s characteristics which distinguish it from other filters include its controllable directional sensitivity and high resistance to noise. The disadvantages of the filter include failure to precisely detect the edges and boundaries of anomalous sources in relation to the total horizontal derivative filter.