Face image resolution enhancement based on weighted fusion of wavelet decomposition

Automatic Border Control (ABC) systems based on face recognition are widely deployed in a border control scenarios. The efficiency of a ABC system depends on the early recognition of the capture subject as soon as he/she appears in the vicinity of the camera. However, due to various factors, including the change in illumination conditions, it is often challenging to capture a good quality face sample to facilitate the speedy biometric recognition with the ABC systems. In this paper, we propose a novel super resolution scheme to enhance the resolution of the captured face image. The proposed method is based on a weighted fusion of high frequency sub-band images obtained using Discrete Wavelet Transform (DWT) and Stationary Wavelet Transform (SWT). The weights are computed and assigned to the corresponding sub-band images of DWT and SWT by measuring both correlation and energy. Then, weighted sub-band images are fused using the sum rule. Finally, all these sub-bands are combined to generate a new resolution enhanced image using Inverse Discrete Wavelet Transform (IDWT). Extensive experiments are carried out using a semi-public ABC database that is comprised of 61 subjects. The database was captured in three different lighting conditions. Further, we also compare the performance of the proposed Super Resolution (SR) scheme with three well established state-of-the-art schemes. The experimental results have indicated that the proposed SR scheme can effectively improve the face recognition component of the ABC system even with low quality face samples.