Research On The Techniques Of Creative Elements Extraction From Textile Images

The pattern and color theme of fabric images can inspire designers. To extract the existing fabric work of art can shorten the creation period of artistic work. Because of the little research of fabric features, it makes the current technology do not suit to handle fabric image. According to the feature of fabric, we propose some methods to adapt previous methods in computer vision to perform geometrical elements and color theme extraction.In the process of pattern extraction, we study the main features that influence geometrical elements extraction, such as fabric noise and continuity. We analyze the formative reasons and results of those features. Then we use mean shift arithmetic and region action graph to smooth the texture noise. In addition, we present the geometric merging strategy to handle objects that coupled appear. The result shows that our method can better keep the integrity of objects than previous related methods.As for color extraction, we mainly analyze the fake colors and the difference between the industrial color model and the model that we used in computer. We employ the salient analysis method to find out which color human focus on. Then we use color quantization to figure out the fake colors. Also we introduce optimization procedure to reduce the gap between two color models. Then the existing database of color theme is used to optimize the color theme. The results show the salient method and color quantization algorithm can successfully deal with the fabric features and the results validate our method.