Human face has its own unique characteristics that are always considered to be the valid biometric feature for automation of security systems. A dedicated system doesn’t only require accurate algorithm but also it is required to be executed with a minimum span of time. In this literature, authors have stated and demonstrated a wavelet based feature space design mechanism from human faces and also analyzed the performance of execution for thermal, digital and range face images. Thus, the developed algorithm for variation of sensors’ (2D and 3D-especially 2.5D face images) data with considerable execution time for any electronic system can be considered for automation purpose. All these face images have been normalized into the same scope, and then Haar wavelet transform have been implemented for them. The simulation model has been designed in MATLAB-SIMULINK software environment, and it is tested for all these three types of image sources. In addition, an array of investigation with different parameter setups has also been reported. According to the analysis of the implemented model, the minimum time span that is required for Haar wavelet transform of 2D visual (UGC-JU database), 2D visual (Frav3D database), and thermal face image is 9.89998 Sec., 9.789 Sec. and 10.299558 Sec. respectively. The same algorithm takes 10.9801 Sec for original 2.5D frontal range face image where for rotated face image, the algorithm at first registers to frontal pose and then performs wavelet transform and takes minimum of 20.77084 Sec.Keywords: 3D face image, 2D face image, Haar wavelet transformation, embedded system, simulation model, 2.5D range image Cite this Article: Suranjan Ganguly, Debotosh Bhattacharjee, Mita Nasipuri. Comparison of Performance of Haar Wavelet Transform on 3D-2D Multi-Sensor Face Images Using Simulation Model. Research & Reviews: A Journal of Embedded System & Applications (RRJoESA). 2015; 3(1): 11–27p.
