Performance analysis of particle swarm optimization algorithm-based parameter tuning for fingerprint image enhancement

Existing algorithms designed for Fingerprint Image Enhancement either lack the ability to enhance poor quality image or are computationally expensive. Evolutionary algorithms are often used to enhance images. Particle Swarm Optimization (PSO) is one of the most progressive algorithms but has parameters, which are not properly tuned to reduce the number of iterations. In this paper, PSO parameters; inertia weight (w) and acceleration constants (c
₁
and c
₂
) were fine-tuned. PSO-based parameterized transformation function, which incorporates both the global and local information of an image was developed to maximize the information content of the fingerprint image. In the transformation function, a threshold of 0.99 was set to control the contrast effect of the enhanced image. The intensity values of pixels that are less than the threshold were transformed. The image quality was evaluated using an Objective Function in term of Number of Edges, Sum of Edge intensities and the exponential of the entropy. The commonly-well-known database FVC-2004 is used in this study. It was observed from the experiments that the best PSO parameters set used for successful convergence of the PSO Algorithm were w ϵ [0.7, 0.75] and (c
₁
, c
₂
) ϵ [1.2,1.3]. Therefore, any set of values used outside these ranges would result to local minimum convergence and increase the computational effort by searching in unwanted areas.