International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B2. Istanbul 2004 
can use mean filter or Gauss filter to filter the additive noise. At 
last, we carry out the exponential operation to the filtered 
logarithmic image. At last, we get the intensity images or 
amplitude images filtered the speckle noise. 
6. The contrast test and Quantitative evaluation 
It is very necessary to appraise the filtering effect. The 
evaluation to filtering speckle noise can be divided into two 
kinds: the qualitative evaluation and the quantitative 
evaluation. The qualitative evaluation can be completed through 
the interpretation with eyes. It is simple and direct, but different 
people will get different qualitative evaluation. So it is 
important to make the quantitative evaluation to avoid one's 
evaluation difference. 
6.1 the ability to restrain the speckle noise: 
It is mainly evaluated from the Filter Index(FI).|Speckle Noise 
Index(0) Equivalent Number of Looks(ENL). 
Filter Index: FI is the ratio of the mean (M) to the standard 
deviation (SD) of homogeneous areas. It can be described as 
followed 
FI -—A£/SD (11) 
The more high FI is, the more strong filter restrains the speckle 
noise. 
Speckle Noise Index(B): It is the ratio of the standard deviation 
(SD) to the mean (M) of the homogeneous areas. 
fE£Y5D/M (12) 
Equivalent number of looks: ENL is another index which 
evaluates the ability to restrain the speckle noise. It can be 
described as follows 
ENL(I) =1/ 8° (13) 
(intensity) 
ENL( A): (0.52277 Ay: (amplitude) (14) 
6.2 Mean tonal value preservation: 
A good filter should preserve the mean backscattering 
coefficient value of homogeneous areas. It is to say that the 
filter should be unbiased estimation. It can be evaluated from 
the normal mean(NM). The normal mean is the ratio of the 
mean of the filtered homogeneous area to the mean of original 
image. It can be described 
up t 
Mean | 
| filtered 
N (15) 
original 
where Mean|originat | Me@Nlfitterea 1S respectively the mean of 
homogeneous area of original image and filtered image. 
Internatior 
Dr 
6.3 Texture and Edge preservation: 
To evaluate the edge preservation, we can use the Edge Keeping 
Index(EKI). The formula of EKI is as follows. 
m m 
EKI 2 G'(w)/ Y G(w) qe 
i=] i=] 
where G( W,) and G'(w,) is respectively donated the 
maximum grey value gradient of original and filtered image in 
the same window. 1-1: m, it donates the number of the sample 
windows. 
The filter window size of the contrast test is 3x3. 
  
The follov 
(c) median filter! 3x3 
Table 1