EXPLORE PUBLICATIONS BY COUNTRIES

Face sketch, one frequency, new dimension, dominant and additional features selection.

Modality reduction by using the Eigentransform method can not efficiently work, when number of training sets larger than image dimension. While modality reduction by using the first derivative negative followed by feature extraction using Two Dimensional Discrete Cosine Transform has limitation, which is feature extraction achieved of face sketch feature is included non-dominant features. We propose to select the image region that contains the dominant features. For each region that contains dominant features will be extracted one frequency by using Two Dimensional-Discrete Cosine Transform. To reduce modality between photographs as training set and face sketches as testing set, we propose to bring the training and testing set toward new dimension by using the first derivative followed by negative process. In order to improve final result on the new dimension, it is necessary to add the testing set pixels by using the difference of photograph average values as training sets and the corresponding face sketches average as testing sets. We employed 100 face sketches as testing and 100 photographs as training set. Experimental results show that maximum recognition is 93%.