针对车载视频图像中同时包含局部运动模糊和全局运动模糊,现有去模糊算法难以适用且效果差等问题,提出一种基于再模糊理论的复杂车载模糊图像复原方法.根据车载视频图像的特点把图像分割为车身和非车身区域,采用改进后的模糊参数估计算法,在车身区域图像块估算出的全局运动模糊参数,对整幅图像进行全局模糊恢复;对复原前后的非车身图像进行分块处理,利用复原前后图像块结构相似度(Structural Similarity,SSIM)和局部均方差的差异性,检测和提取出局部模糊区域;对提取的模糊区域进行复原后与清晰区域拼接融合,合成清晰的图像.与现有算法对比实验分析,所提算法取得了不错的效果,且复原后图像的峰值信噪比(Peak Signal to Noise Ratio,PSNR)和SSIM表现良好.
Vehicle-mounted Complex Image Restoration Based on Improved PSF Estimation
Both local motion blur and global motion blur are included in vehicle video images.The existing deblurring algorithms are difficult to apply and the effect is poor.A complex vehicle blur image restoration method is proposed based on re-blur theory.Firstly,according to the characteristics of the vehicle video image,the image is divided into the body and non-body areas.The improved blur parameter estimation algorithm is adopted to estimate the global motion blur parameters in the image block of the body area and perform global blur restoration for the whole image.Then,the non-body images before and after restoration are blocked,and the local blurred areas are detected and extracted by using the difference between Structural Similarity(SSIM)and local mean square deviation before and after restoration.After the extracted blurred area is restored,it is stitched and fused with the clear area to synthesize a clear image.Compared with the existing algorithm,the proposed algorithm achieves good results,and the restored image performs well in Peak Signal to Noise Ratio(PSNR)and SSIM.
万方数据
维普
