基于小波变换的纹理合成算法研究

Algorithm Researching of Texture Synthesis Based on Wavelet Transform

基于样图的纹理合成基本思想是根据给定的小区域的纹理样本,拼合生成任意大小的纹理图像并且在视觉上与样本是相似而连续的。通过对基于样图的纹理合成算法进行研究,在待合成纹理块的搜索策略上,提出优先搜索前一个已合成纹理块在样图中的位置附近,快速定位下一个用于合成的纹理块。在纹理块的匹配策略中引入小波变换,结合小波变换后的小波系数和小波变换后的低频域图像进行块的重叠区域匹配,最后使用最小路径法将待合成纹理块缝合到输出图像中。充分考虑了样图的结构信息并提高了合成速度,获得了较好的纹理合成效果。

Basic idea of the algorithm of texture synthesis form samples was to collage large texture images from samples. The algorithm of texture synthesis from samples was studied in this paper and a new algorithm was proposed. When searching for undetermined textural blocks, locations neared the last determined textural block in sample image would be searched firstly, so the undetermined textural block can be determined quickly. Wavelet transform was used for matching; the wavelet coefficients and low-frequency domain image of wavelet transform were used together in texture matching process. At last, the undetermined texture block was collaged by minimal path method. Structural information was considered fully, the process was quickly and the texture synthesis results were satisfactory.