Skip to Main Content
Massive amount of data with high dimensionality can pose a problem for efficient image classification. Recently there has been an effort to extend the application of sparse representations of signals to image classification. In this paper, we propose a method that extracts the smallest number of features that discriminate the images from different classes using a cost function that combines discrimination power and sparsity. The proposed method was evaluated using the TU Darmstadt database and was compared with Linear Discriminant Analysis (LDA) and was shown to achieve higher accuracy with smaller number of features than LDA. The robustness of our method to noise and occlusion was also illustrated through experiments.
Statistical Signal Processing Workshop (SSP), 2012 IEEE
Date of Conference: 5-8 Aug. 2012