We present a novel algorithm, namely Feature Tracking Velocimetry (FTV), which is less sensitive to the appearance and disappearance of particles and to high velocity gradients than classical Particle Image Velocimetry (PIV). The basic idea of FTV is to compare windows only where the motion detection may be successful, that is where there are high luminosity gradients. The FTV algorithm is suitable in presence of different seeding densities, where other techniques produce significant errors, due to the non-homogeneous seeding at the boundary of a flow. The FTV algorithm has been tested for the analysis of laboratory experiments on simple jets (SJs) and negatively buoyant jets (NBJs), both issuing from a sharp-edged orifice. Among the others, the velocity and Turbulent Kinetic Energy profiles, orthogonal to the jet axis, the mean streamwise centerline velocity decay and the integral Turbulent Kinetic Energy along the jet axis have been measured and analyzed. These quantities have been employed to study the differences between simple jets and NBJs, and to investigate how the increase in buoyancy affects the NBJ behavior. Moreover, mean velocity fields have been used to study the geometrical dimensions of the jet, while second order statistics, such as Turbulent Kinetic Energy, have been analyzed to characterize the turbulence structure governing the mixing processes.