![]() | يوجد فقط 14 صفحة متاحة للعرض العام |
المستخلص Since the trend for automatic based coloring is an imporgant goal for many image processing researchers especially in this field, we have proposed in an intelligent fully automatic coloring system for textural image like natural images. It works by segmanting the image into different textured regions then finding the suitable for each region from a set of different textures stored in a special database and then recoloring the image parts by the attached color with the recognized class. The texture features used in this paper have varid between wavelets coefficients, Laws kernels, Co-occurrence matrix and Tamoura measures. Both mean shift and Fast k-Mean clustering techniques were used in the segmentaion stage. kNN classifier is used in the classification stage. Finnaly, the coloring process was done using HSV color space to preserve reality in the image. Also we have proposed simple modofocation to improve our system results and to make fully automatic system. |