| |
| Acadêmico(a): Rafael Weimer Baade |
| Título: DERMANOSTIC: um método para normalização da iluminação em imagens de lesões cutâneas |
| |
| Abstract: |
| This work presents an implementation of a method for the normalization of the lighting in skin lesions images captured by standard cameras. The normalization is realized by transforming the color system of the image to Hue Saturation Value (HSV). The algorithm uses the V color channel, because it’s the channel where the illumination remains better represented. From this channel a sample matrix of 400 pixels is extracted from the four corners of the image. The sampling is used to generate the mathematical model of natural lightening. This model is used to generate new values for the channel V through the division of the current value of V for the value obtained by the mathematical model. To return the color to the natural tonality it’s realized a multiplication of the new values of channel V through the average values of the original channel V and the division for the average of the new values. Lastly the image is converted to the original color system. To verify the efficiency of the lightening normalization method, the method was tested with the segmentation algorithm used by Pradi (2012). The results presented an improve of 7,93% in the segmentation results. |
| |
|
