Improvement Of Plate Character Image Defimition Based On Image Analogies

Degraded images of license plate characters, not only is difficult to meet humanperception, but also create problems for latter identification, analysis and processing. So,to ameliorate the quality of the plate images and to facilitate the viewing of platecharacter images are of important significance.Image analogies algorithm does perfect job in art filtered mirror, texture synthesisand so on. IA algorithm is also used in super-resolution technology. Based on imageanalogies, an algorithm is proposed to improve the plate character image definition.According to the target character image, training set is structured and to learn thepotential relationship between the pair of images in the training set, quantify the humansenses as the function mapping and then use the relationship to synthesize the outputimage.The experimental results of simulated and real character images taken by cellphone camera are shown below. This method can generate more reasonable characterdetails than general image analogies method. In the simulated image experiments,compared with conventional filters, the new method can reduce the noise, andeffectively improve the brightness of low illumination image. In the real imageexperiments, it can extract more right SIFT feature matching points than target imageand make for later pattern recognition and analysis. For the almost unrecognizedcharacter images, this method can determine the character and lead to better visualeffects. The performance validation of new method proved that it could remove noise,improve visual effects in the low illumination images, and give a nearly perfectcharacter image.