Extraction of linear features from Remote Sensing Image (RSI) has found many applications as in urban planning, disaster mitigation and environmental monitoring. There were many previous studies in this field appreciating the significance of statistical operators to extract linear features. But in RSI domain, it has a different significance as it involve handling a large data set of multiband data involving complexities in terms of spectral, spatial and temporal domain. Most of the objects in nature were not easily discernable and extracted as they were often contaminated or mixed with other objects and might influence the spectral character of the object. This may be less in urban environment as they exhibit more or less uniform spectral behavior where as in natural setting it may exhibit complex spectral behavior. Present study demonstrates such complexities in extracting linear features in different setting – urban and coastal area – using first order derivative gradient filters.