RESEARCH ON GRIDDING PROCESSING STRATEGIES OF REMOTE SENSING 
IMAGE SEGMENTATION BY REGION GROWTH 
ZHU Hong-chun®®, ZHANG Ji-xian®, LI Hai-tao®, YANG Jing-hui®, LIU Hai-ying® 
©Chinese Academy of Surveying and Mapping, Beijing 100039, China; (2)Geo-information Science & Engineering College, 
Shandong University of Science and Technology, Qingdao 266510, China; ©The Department of Basic Courses, Shandong 
University of Science and Technology, Qingdao 266510, China 
KEY WORDS: region growth, image segmentation, gridding, parallelism 
ABSTRACT: 
Region growth is an important method based on regional information, and its gridding process also is an important problem. This 
paper starts with basic principle and mathematical models of region growth. Then some new serial region growth algorithms 
proposed in recent years are classified and analyzed, based on which their related gridding strategies are emphatically discussed. 
Then we analyze and put forward the algorithm, which should be feasible when designing gridding parallel segmentation, and give 
an evaluation of the parallel regional segment algorithm. Finally, we point out the problems and challenges of future research. 
1. INTRODUCTION 
Remote sensing image segmentation is the first step towards 
image analysis and application, and it’s the important 
components of image understanding. The so-called remote 
sensing image segmentation, based on the objectives and 
background of a priori knowledge, of the mean is marked target 
image and positioned background, and it is said to be physically 
meaningful regional connectivity collection of remote sensing 
for further lay the foundation for image classification. 
Segmentation final result of the image is decomposed into a 
number of characteristics with the minimum components. There 
are many major traditional segmentation algorithms such as 
regional and border detection etc. Faced with geometrically 
growth of remote sensing data and rapid treatment needs, a 
more efficient algorithm and strategy for image segmentation is 
cried for. 
Grid has become the third generation of the Internet, promoting 
a higher level of resource sharing. In grid environment, all 
kinds of special scientific computing and processing can get 
maximize the efficiency and speed. Therefore, the study for the 
development of Grid remote sensing image processing method 
is a practical significance of the issue. This study focuses on the 
regional growth of remote sensing image segmentation methods 
of Grid strategy and algorithm, and evaluation of the capability 
is put up based experimental. 
2. REGIONAL GROWTH SEGMENTATION 
DEFINITION AND THE TRADITIONAL 
ALGORITHM 
2.1 Regional growth definition 
Regional growth is an image segmentation method that is very 
concerned about by the Computer Vision, and it is particularly 
suited to image texture segmentation. The basic idea is to form 
a region through together similar pixels. Input image is often 
firstly to be divided into a number similar region, and then 
selected a seed and iterative followed with seed pixels around 
similar pixels according to some criterion. One its’ focus is the 
first seed pixels choice and the growth of design rules the other 
is the efficient and accurate algorithm. 
2.2 Regional growth algorithm 
Step 1: Through image scanning, the seed is identified to 
regional growing; 
Step 2: Seed pixels’ gray or band characteristics compare with 
other surrounding pixels’ (four or eight neighbourhood). If meet 
the established criteria for the judgment, then fused as a region; 
Step 3: For newly merged pixels, repeatedly step on the 
operation; 
Step 4: Repeated two, three steps, not until the regional 
expansion; 
Step 5: Returning to the step 1, and searching new growth point 
in the region of pixels. 
In general growth algorithm, there are two key points. One is 
the seed selection the other is the formulation of guidelines for 
growth. Seed pixels selections mainly include the following 
methods: 
(1) Using iterative methods to select the smallest contraction 
seed, the shortcomings of this method is the calculation of the 
additional burden; 
(2) Because the targets mostly have the more larger radiation, 
so we can choose the most image-pixel as seed. 
(3) If there is no priori knowledge of division, we can calculate 
each cluster using the growth criteria. And if the calculation 
results showed clustering, the pixel that close to the clustering 
center is chosen as seed pixels. 
The main criteria for growth based on the following 
characteristics: 
(4) Through Comparing seed gray difference, the difference is 
seemed as a gray Threshold. Identified the threshold method is 
based on the mean tonsure or the concavity based on the gray 
histogram analysis. 
(5) The cumulative histogram can be used, which determining 
regional integration using the rule that gray distribution is 
similar. 
(6) The shape characteristics of the region Segmentation can be 
use as the decision to terminate the growth conditions, but also 
as a regional merger conditions.