摘要
对于任意运动模糊方向和长度的模糊图像复原的关键在于点扩展函数的准确性,点扩展函数的精度由模糊方向和长度决定.本文首先基于Hough变换在模糊图像频谱上对运动方向和长度进行初次识别,然后在以初次识别结果为中心的一个较小邻域内基于维纳滤波进行复原,得到一系列随模糊长度和方向变化的复原图像,以复原图像和模糊图像的峰值信噪比为评价指标,取最大信噪比值的图像为最佳复原图像,其对应长度和方向为最终识别值.仿真实验表明,本文识别结果精度非常高,复原效果较为理想,算法原理较为简单实用,模糊方向绝对误差不超过1°,模糊长度绝对误差不超过1像素.
For any motion blurred image of the direction and length,the key to recover it is the accuracy of point spread function. The precision of point spread function depends on blurred direction and length. This paper initial i- dentified the movement direction and length based on Hough shift in blurred image spectrum; then in view of the results as a center in a smaller area based on wiener filter to restore blurred image to obtain a series of restored images that change with fuzzy length and direction. Taking peak signal noise ratio of restored image and blurred image for evaluation indexes to make the best restored image which corresponds to maximum peak val- ue signal to noise ratio evaluation index image, its corresponding length and direction is its final recognition value. The simulation experiments proved that this paper＇s recognition results has high precision. The effect of the restored image are ideal, the principle of this algorithm is simple and practical; the absolute error of blur direction does not exceed 1 degree; and the absolute error of blur length does not exceed 1 pixel.
出处
《三峡大学学报（自然科学版）》
CAS
2012年第6期93-97,102,共6页
Journal of China Three Gorges University：Natural Sciences
关键词
运动模糊图像
频谱
HOUGH变换
峰值信噪比
图像复原
motion-blurred image
frequency spectrum
Hough transformation
peak signal niose ratio
image restoration
