运动模糊图像检测分割盲复原

发布时间：2019-05-30 08:25

【摘要】：针对空间变化的运动模糊图像,提出基于单帧图像的模糊检测与分割的盲复原方法。首先,模糊检测生成模糊区域映射图;其次,采用腐蚀和膨胀分别对映射图中的模糊区域与清晰区域处理,利用处理结果标记,并采用分割算法分割图像;再次,分两步估计模糊核,并对模糊区域进行外延扩张,利用估计得到的模糊核进行反卷积去模糊;最后,去除被外延的部分,并与分割得到的清晰区域进行拼接,得到完整的去模糊后的图像。结果表明,算法在人工数据集峰值信噪比(peak signal to noise ratio,PSNR)平均提高10dB,均方根误差(root mean square error,RMSE)平均降低0.12,自然场景下的真实模糊数据集去模糊结果视觉质量明显提高。此外,该算法在减少模型复杂度的同时也提高了去模糊的质量。
[Abstract]:Aiming at the motion blur image with spatial change, a blind recovery method based on fuzzy detection and segmentation of single frame image is proposed. Firstly, fuzzy detection is used to generate fuzzy region mapping map, secondly, corrosion and expansion are used to deal with fuzzy region and clear region respectively, the processing result mark is used, and the segmentation algorithm is used to segment the image. Thirdly, the fuzzy kernel is estimated in two steps, and the extension of the fuzzy region is carried out, and the fuzzy kernel obtained by the estimation is used to de-blur the fuzzy kernel. Finally, the extended part is removed and spliced with the segmented clear area to get the complete deblurring image. The results show that the peak signal-to-noise ratio (peak signal to noise ratio,PSNR) of the algorithm in artificial dataset is increased by 10 dB on average, and the root mean square error (root mean square error,RMSE) is reduced by 0.12 on average. The visual quality of the real fuzzy data set deblurring results in natural scenes is obviously improved. In addition, the algorithm not only reduces the complexity of the model, but also improves the quality of deblurring.
【作者单位】：北京交通大学信息科学研究所;现代信息科学与网络技术北京市重点实验室;
【基金】：国家自然科学基金(61572063)资助课题
【分类号】：TP391.41

【相似文献】