A Hybrid Median Filter for Enhancing Dim Small Point Targets and Its Fast Implementation

Weak stars and middle altitude orbit satellites or high altitude orbit satellites are presented as dim point targets in images of astronomical observation. They have low intensity and occupy few pixels in image. In most situations, the background of sky is always ununiform and there exists pulse noises. Thus results in targets submerging in background noises and low signal noise ratio $(SNR)$. Median filter is an effective method for removing pulse noise. But it can blur image after modifying gray of non-noise pixels when image has low SNR. A hybrid median filter algorithm is proposed to process low SNR dim point targets image. The newly proposed method can suppress the pulse noises, also can increase target $SNR$ and does not add extra fake targets. The time complexity of median filter is $O(N^2)$($N$ is the widow size) and is difficult to implement in hardware. For quasi real time application, a fast implementation of our hybrid median filter is proposed to reduce the time complexity to $O(N)$. Experimental results show that the newly proposed algorithm is superior to ordinary median filter and fast median filter by comparing visual effect and performance.