CAJ | 학술논문

为解决图像中运动物体的模糊问题，提出一个基于图像视觉显著性的快速去模糊机制。根据图像的显著图将其划分为视觉显著区域和非显著区域，使用改进的快速去模糊算法对非显著区域的模糊核进行预测；使用可以将显著区域运动物体的空间可变模糊转换为空间不可变模糊的补偿机制对显著区域进行补偿，把预测到的非显著区域模糊核应用到显著区域的去模糊处理中；使用反卷积算法恢复图像。实验结果表明，在保证整个图像具有统一的模糊度和不产生回铃现象的前提下，该机制具有较好的去模糊效果。
위해결도상중운동물체적모호문제，제출일개기우도상시각현저성적쾌속거모호궤제。근거도상적현저도장기화분위시각현저구역화비현저구역，사용개진적쾌속거모호산법대비현저구역적모호핵진행예측；사용가이장현저구역운동물체적공간가변모호전환위공간불가변모호적보상궤제대현저구역진행보상，파예측도적비현저구역모호핵응용도현저구역적거모호처리중；사용반권적산법회복도상。실험결과표명，재보증정개도상구유통일적모호도화불산생회령현상적전제하，해궤제구유교호적거모호효과。
Aiming at the blur of the fast moving object in video images ,a new fast deblurring scheme base on the visual saliency was presented .Firstly ,the image was divided into a significant part and a non-significant part through its saliency map ,the blur kernel of the non-significant regional was estimated with an improved fast motion deblurring algorithm .Then ,a compensation mechanism was proposed to convert the variable space blur of the significant part into the invariable space blur ,and the estimated blur kernel mentioned before was applied to deblur the significant part .At last ,the image was restored using the deconvolution algorithm .Experimental results show that the scheme can reduce the running time of the image deblurring process with the uni-form blur and maintain the deblurring quality on the motion object .