Pyramidal Watershed Segmentation Algorithm
for High-Resolution Remote Sensing Images Using
Discrete Wavelet Transforms

1. Introduction

Image segmentation is object oriented and hence useful in high-resolution image analysis. Itprovides a partitioning of the image into isolated regions, each one representing a differentimage. In particular, multiscale methods are based on image transformations that reduceimage resolution, and they are able to segment objects of different sizes depending on Page 2 2Discrete Dynamics in Nature and Societythe chosen resolution. From high-resolution images, small details can be detected, whilefrom low-resolution images, larger structures are detected. The multi-scale behavior ofimage features has been analyzed in different ways. Tracking of intensity extrema alongscales defined by Lifshitz 1 and Lindeberg 2 was used for image segmentation. Otherapproach for image segmentation is based on the watershed transform. This transformcan be applied to the gradient magnitude image as defined by Meyer and Beucher 3 ,Vincent and Soille 4 , and Bieniek and Moga 5 to obtain the segmented regions. Smallfluctuations in the grey levels produce spurious gradients, which cause over segmentation.To overcome this problem many techniques based on watersheds have been proposed. Meyer 6 introduced the leveling approach, which applies morphological filters to reduce the small details in the image. Jung and Scharcanski 7 , Jung 8 and Haris et al. 9 proposededge preserving statistical noise reduction approach as a preprocessing for the watershedtransform, and a hierarchical merging process as a postprocessing stage. Jackway 10 andGauch 11 used morphological scale spaces for image segmentation. Weickert 12 proposedpartial differential equations for image denoising and edge enhancement, combined withwatershed segmentation and region merging. J. B. Kim and H. J. Kim 13, 14 proposed awavelet-based watershed segmentation technique, by projecting the segmented image intohigher resolutions. Nguyen et al. 15 proposed a combination of energy-based segmentationand watersheds, called water snakes. Recently Liu et al. 16 modified the wavelet-basedwatershed segmentation algorithm, introduced the morphological filters to smooth imagewhile filtering out noise as preprocessing of medical image segmentation. Almost all theproposed techniques are applied on the medical and nonmedical images

2. Methodology

In this paper, we proposed a new segmentation algorithm for high-resolution remote sensingimages, which can also be applied to medical and nonmedical images. We used a bi-orthogonal bior wavelet decomposition to describe a remote sensing image in multipleresolutions. A suitable resolution is chosen. The gradient image is estimated or computedby the simple grey scale morphology. To avoid over segmentation, we have imposed theselective minima regional minima of the image on the gradient image. The watershedtransform is applied and the segmentation result is projected to a higher resolution, usingthe inverse wavelet transform until the full resolution of segmented image is obtained. Ageneral outline of the proposed method is shown in Figure 1.

The results of wavelet transform are many wavelet coefficients C, which are functions ofscale and position, s t is the image function and ψ scale, position,t is the wavelet function.If scales and positions based on powers of two, the so-called dyadic scales and positions thenour analysis will be more efficient and just as accurate. We can obtain such analysis from theDiscrete Wavelet Transform DWT . For many images, the low frequency content is the mostimportant part. This is the one which gives the image identity. The high frequency contentimparts flavor or nuance.According to Mallat’s pyramidal algorithm 17 , the original image is convolvedwith low-pass and high-pass filters associated with a mother wavelet, and down sampledafterwards as shown in Figure 2. In order to explain the process of study, a satellite image ofVisakhapatnam city coastal harbor area has been chosen. It is PAN sensor with 5.8 m spatialresolution of IRS-1C satellite and it is as shown in Figure 3

3. Experimental Results

All the computation involving wavelet transform, edge detection and watershed transformare implemented using MATLAB. Application of the wavelet transform takes very shorttime, this quick response is mainly due to the property of wavelet decomposition andreconstruction which have fast algorithms. The algorithms are based on convolutions witha bank of filters. We used two different satellite images. Satellite image 1 is a PAN imageof IRC-1C 5.8 m resolution . It shows Visakapatnam sea coast with fishing jetties, harborchannels and dense city with concrete structures. Image 2 is a Cartosat-1 satellite PAN image2.5 m resolution of Hyderabad city. The final segmentation results are shown in Figure 6.The Figure 6 a are the original images and Figures 6 b to 6 d are the watershed imposedoriginal images from level 2 to level 0, respectively. At the lower resolution level 2 , weselected the minima to detect minimum number of intensity valleys, which gives less number

4. Conclusions

In this paper, we proposed a new technique to improve the watershed segmentation onremote sensing images based on Discrete Wavelet Transforms. The lower resolution mustbe chosen in accordance with the size and number of desired objects to avoid undersegmentation. The experimental results indicate that the proposed technique performs wellfor the remote sensing images as well as for medical and low-resolution images too. The fullresolution image has more number of regions, longer computation time and lowest minimavalue where as the low-resolution image has less number of regions, shorter computationtime and higher minima value. We would like to extend this work using other soft computingtechniques.

Acknowledgment

The authors thank the All India Council of Technical Education for utilizing the data obtainedin the research project for this paper. The authors also thank Andhra Pradesh State RemoteSensing Application Center for providing the high-resolution satellite data. The medical andnonmedical images are downloaded from Google.com/images

Полная версия статьи доступна по следующей ссылке:

http://www.hindawi.com/journals/ddns/2009/601638.html