Integrating the Maximum Line Gradient Value Improvement and the Level Set Method Variation to Detect Face Sketch Multi-Features

Deformable model is algorithm used to represent non-rigid object, but only a few researchers that work using different modality, where photograph images as training set and sketch images as testing set. “Face sketch multi features detection based on the line gradient and the adaptive shape variants average values method” is one example of deformable model. It has the weakness, incapacity to move shape toward its corresponding feature, when shape lies on the same gradation. In this paper, we proposed method to simultaneously move shape, when shape can not move on the same gradation. To overcome difference of scaling, translation on testing set, we have enhanced landmark initialization by adjustment scaling and translation of training set as landmark initialization. Shape will be moved based on integrating maximum line gradient improvement and the level set method variation. We have employed 100 face photographs as training set and 100 hatching face sketches as testing set. For each testing set, we tested by using 7 shapes of 28 landmarks. The experimental result shows that percentage of detection rate is 89.79%.