With the increase of on-orbit manipulations and space conflictions, missions such as tracking and capturing the target
spacecrafts are aroused. Unlike cooperative spacecrafts, fixing beacons or any other marks on the targets is impossible.
Due to the unknown shape and geometry features of non-cooperative spacecraft, in order to localize the target and obtain
the latitude, we need to segment the target image and recognize the target from the background. The data and errors
during the following procedures such as feature extraction and matching can also be reduced. Multi-resolution analysis
of wavelet theory reflects human beings' recognition towards images from low resolution to high resolution. In addition,
spacecraft is the only man-made object in the image compared to the natural background and the differences will be
certainly observed between the fractal dimensions of target and background. Combined wavelet transform and fractal
dimension, in this paper, we proposed a new segmentation algorithm for the images which contains complicated
background such as the universe and planet surfaces. At first, Daubechies wavelet basis is applied to decompose the
image in both x axis and y axis, thus obtain four sub-images. Then, calculate the fractal dimensions in four sub-images
using different methods; after analyzed the results of fractal dimensions in sub-images, we choose Differential Box
Counting in low resolution image as the principle to segment the texture which has the greatest divergences between
different sub-images. This paper also presents the results of experiments by using the algorithm above. It is demonstrated
that an accurate texture segmentation result can be obtained using the proposed technique.