| Abstrakt: |
Recently, the effect of sun exposure has become more harmful than ever before. People begin to worry about the health of their skin, especially facial skin. Dermatologists and beauty clinics try to make an instrument that can help people in analyzing facial conditions. However, the instruments are often costly and not flexible for ordinary people. For example, to analyze our facial skin condition using the instrument, we have to come to a beauty clinic that has the instrument and needs expert's help to use it. Special settings for image acquisition are also needed, which make the self-assessment quite difficult to perform. Therefore, this study attempted to build a prototype system that can analyze facial skin conditions automatically without rigid instructions and can be performed using a mobile camera without involving expensive instruments. In the initial stage, we focused on developing a prototype for automatically identifying the presence of acnes and their position in the face. Data used to test the reliability of the prototype are close-up face images that have acnes, collected randomly from the Internet. We employed the following image processing technique to process acnes' detection: image resizing, image conversion, image segmentation, image sharpening, and feature extraction, respectively. For evaluation, the output image from the system was compared to the consensus marking from dermatologists. Comparative results were assessed using sensitivity, specificity, and accuracy. The experiment result shows that average sensitivity, specificity, and accuracy values from all images used are 0.47, 0.97, and 0.92, respectively. Based on these scores, it can be interpreted that the system has the potential to detect non-acne objects well, but it requires improvement for the sensitivity score. [ABSTRACT FROM AUTHOR] |
