The Superconducting Quantum Interference Device (squid) sensor is the most sensitive device available to detect magnetic flux.Due to its high sensitivity, SQUIDs are often coupled to gradiometers - a set superconducting coils connect in a subtractive way - in order to reduce environmental noise. Because the response of the sensor depends on the gradiometer geometry it is of great importance the development of an analytical model that accurately reproduces the gradiometer effects over the output signal and also helps in the design of such devices. A novel model based on signal processing methods is proposed in this Thesis, which considers gradiometers as multi-dimensional spatial filters with the magnetic field as input and the net magnetic flux as output. The main configurations found in the literature are compared and commented, and gradiometers are designed to detect magnetic field in a typical application of SQUID sensors.Moreover,it is studie the decovolution issue, which allows the obtainment of the magnetic field( what is to be measured in most cases )from flux measuresmentes (what is actully measured by SQUID sensors.