CAJ | 학술논문

针对遥现系统中，由于相机精度、材质等原因导致深度信息质量不高，存在缺失以及噪声等问题，提出了基于显著特征的滤波算法。结合频率域的显著图分析算法和基于图的分割算法提取出人类感兴趣的显著区域，针对这部分区域利用联合双边滤波法进行重点处理。实验结果表明，该算法在有效抑制了平坦区域的噪声的同时，增强了显著区域深度图像的细节，并填补了边缘部分的缺失，实现了实时条件下深度图像的有效滤波。
침대요현계통중，유우상궤정도、재질등원인도치심도신식질량불고，존재결실이급조성등문제，제출료기우현저특정적려파산법。결합빈솔역적현저도분석산법화기우도적분할산법제취출인류감흥취적현저구역，침대저부분구역이용연합쌍변려파법진행중점처리。실험결과표명，해산법재유효억제료평탄구역적조성적동시，증강료현저구역심도도상적세절，병전보료변연부분적결실，실현료실시조건하심도도상적유효려파。
In telepresence there are many critical problems, such as low quality depth data, absence of depth information, and noise, which are caused by the precision and material of camera.For solving these problems, a salient features based filter algorithm is proposed.The algorithm first extracts a salient region by the frequen-cy-tuned method and graph-based image segmentation method.Next, the algorithm uses joint bilateral filter to process this region on purpose.Experimental results show that the proposed method can effectively reduce noise, recover the lost region and enhance details at the same time.This allows for achieving filtering depth map effectively in real time.