Research on an Extracting Method to Assist Manual Identification of Targets - Advances in Image and Graphics Technologies - page 116

Fig. 1. Resolution Reduction Fig. 2. Image Segmentation

Fig. 3. Linear Gray Level Transformation Fig. 4. Histogram Equalization

From the Fig. 5, one feature can be found out: that is objects are white inside basi-
cally and connect into blocks while the pixels in background area are randomly black
or white. The white point in the background area is scattered and can not be linked up
into larger blocks. Erosion is used to decrease the connection degree of white points
in the background area further. The erosion template, the size of which is

3 × 3 , is

used to process image. The result is Fig. 6.
In Fig. 6, the area of lawn is similar to salt-and-pepper noise and the suspected tar-
gets occupy the large white blocks. The above characteristic can be used to eliminate
useless areas and reserve suspected targets. On this picture, take 10 as N , which is
the number of blocks to be reserved.
After the observation of Fig. 7, one can realize that suspected targets have been ex-
tracted out. But there is some lack of pixels and outline is not obvious. The dilation
processing can make the white area expanded and diminish the hole inside the target.
So, template

3 × 3 is used to process the image and Fig. 8 is the result.