A Study of Local Adaptive Gradient Median Filter

국부 적응 변화율 메디안 필터에 관한 연구

Two-dimensional median filters were affectively supprssing the noise in image processing with the adge smearing decreased. However, if the window were large as necessary in noise then the filter had tendency to cut off corners. An estimate of gradient was used to decide how the ouputs of the filters were calculated. For parallel to the gradient direction we used edge preserving median operation and orthogonal to that averaging subfilters over which medians were then chosen. Four different algorithms were introduced.

제안된 국부적응 2차원 변화율 메디안 필터는 변화율의 추정치에 의해 필터를 출력 방법을 결정하여 경계를 손상시키지 않고 잡음을 제거하는 효과를 가진 4가지의 알고리즘으로서 Subfilter Size를 변형하면 Corner를 복원할 수 있고 평균 Subfilter 대신 중앙연상르 취하면 완벽한 Cornoer 복원이 된다. 실 영상에 적용하여 알고리즘의 성능이 경계를 흐리지 않고 잡음의 영향을 감소하는 효과를 확인하였다.