基于模糊度参数估计的数字媒体图像无损复原

Lossless restoration of digital media images based on ambiguity parameter estimation

传统的数字图像修复方法在实际拍摄曝光时间内,相机与拍摄对象的相对位移易产生误差,导致复原结果较差。为此,提出基于模糊度参数估计的数字媒体图像无损复原方法。构建模糊图像退化模型,测量出频谱图中相对运动方向和水平方向夹角,估计运动模糊方向,采用空间笛卡尔坐标系对运动模糊度参数进行估计。通过径向基函数(RBF,radial basis function)神经网络结构,计算输入函数,将估计结果与实际拍摄曝光时间内相机与拍摄对象相对位移比较,计算出模糊度误差。采用维纳滤波复原法设计复原流程,实现数字媒体图像无损复原。实验结果表明,所提方法的模糊方向容错控制在1°以内,模糊度容错控制在0.2以内,模糊度改善在40%以上,图像复原结果较为清晰。

Traditional digital image restoration methods produce errors in the relative displacement between the camera and the subject during the actual shooting exposure time, resulting in poor restoration results. A lossless restoration method based on ambiguity parameter estimation is proposed. An ambiguity image degradation model is built to measure the relative motion direction and horizontal direction angle in the spectrum, estimate the motion ambiguity direction, and the motion ambiguity parameter by using the space Cartesian coordinate system. Through the radial basis function neural network structure, the input function is calculated, and the relative displacement of the camera and the shooting object during the actual shooting exposure time is compared with the estimated result, and the ambiguity error is calculated. Wiener filter restoration method is used to design the restoration process and realize the lossless restoration of digital media images. According to the experimental results, the fault tolerance of ambiguity direction of the proposed method is controlled within 1°, the fault tolerance of ambiguity degree is controlled within 0.2, and the improvement percentage of ambiguity degree is more than 40%. The restored images are relatively clear.