Elastic Net for solving sparse representation of face image super-resolution

Super-resolution is very important in recognizing suspects face in video surveillance system. In this paper, we present an improvement of image super-resolution based on sparse signal representation. The issue of how to deal efficiently with sparse feature has great significance on the quality improvement of generated high resolution image. We propose to use Elastic net to solve sparse representation of image super-resolution process. Elastic Net will compromise between Lasso and Ridge regression to find the best correlated patch between low-resolution and high-resolution image. Experiments demonstrate that using Elastic Net reconstructed high-resolution images have better color and texture quality. It also gives smaller value of Root Mean Square Error (RMSE) than Lasso and other conventional methods. Small RMSE means more accurate when recognizing face.