基于纹理合成算法在恢复污损牌照中的应用

车牌识别系统是现在智能交通系统中信息采集的一种手段，但是由于图像模糊、字符污损、粘连等实际问题导致识别的瓶颈。文中提出了一种基于纹理合成的图像修补算法——二维表面纹理填洞算法。该算法是根据区域生长得到车牌字符污损部分的边界，即算法中的“洞”，再基于纹理片合成与Quilting算法相结合提出了一种二维表面纹理填洞算法，针对“洞”进行图像纹理修补。实验结果表明，该算法对恢复字符的完整性和修复纹理具有很强的自适应能力，极大地提高了后续的车牌字符分割和识别率。

Removing Color Stain on Vehicle License Plates Based on Texture Synthesis Algorithm

Although recognition of vehicle license plate is well studied,characters can be obscured by color stains due to rust,mud,peeling paint or fading colors and become unrecognizable.Puts forward a hole-filling algorithm of 2D surface texture based on texture synthesis.Firstly,the edge of the absent part of the feculent car's license plate can be achieved based on region growing technology,which is the "hole" in this method.Secondly,on the basis of texture synthesis and Quilting method,advances a hole-filling method of 2D surface texture.Last,aimed at the "hole" the image texture can be repaired.This method is proved to be effective to clean the blurry license plate images,which provide a better input for the successive processes,such as character segmentation and recognition.