3rd International Conference on Electrical & Computer Engineering ICECE 2004, 28-30 December 2004, Dhaka, Bangladesh ISBN 984-32-1804-4 96 AN ALGORITHM TO FIND THE AREA OF IMAGE OBJECTS USING POLYGONAL APPROXIMATION M. Ashraful Amin 1 , Shamim Akhter 2 , Nitin V. Afzulpurkar 3 1 Department of Computer Science and Information Management, Asian Institute of Technology, Thailand. 2 Department of Computer Science, American International University of Bangladesh, Bangladesh. 3 Department of Industrial System Engineering, Asian Institute of Technology, Thailand. E-mail: amin021us@yahoo.com, shamim@aiub.edu, nitin@ait.ac.th ABSTRACT One of the important issues in the image feature is the area of objects in the image. Here we propose a method to calculate the area of an image object. This method finds the polygonal area approximation of image objects. A sequential fine-tuning of this polygon is used to estimate the actual area. At first, we present an algorithm to estimate the object area from Convex-Hull approximation. Finally, a polygonal approximation is used to find the object's area to reduce the Convex-Hull calculation overhead. 1. INTRODUCTION Different types of feature [1, 2] descriptions are in practice in image processing field. One of the common features is the ‘Area’ of image objects. Convex-Hull is an approximation of image area and feature. Here we propose an algorithm that will find more precise area of image objects from Convex- Hull approximation. Finally, a faster algorithm than the initial one is proposed. Where we discuss how it is possible to find the precise area of an object from an arbitrary polygonal approximation. This algorithm is inspired by some common image processing algorithms such as priory boundary detection [5] and Convex-Hull [8].There are several algorithms available for detecting the boundary of an image object. Some of them are shaped in the form of morphological operators [1], whereas others are based on boundary refining or boundary growing [5]. We will discuss two algorithms; concept of these algorithms is used in the proposed algorithms. 1.1 Boundary refining & boundary detection We focus on the approximate boundary estimating algorithms. And utilize these algorithms to find more precise area to contribute in feature calculation. There are two algorithms that can be used consecutively to grow a better boundary of an image object. The First algorithm is to find the likely position of the boundary of the image object. The priory boundary detection algorithm will find points on the boundary in intervals (depending on some understanding of the concerned object). Figure-1 shows a situation where prior knowledge of the approximate center (o) of the object is used. If we lack information about the object in hand then we can scan from left to right and then find the other points of boundary in a fixed period or interval (Figure-1). Here we present two enlarged sections of the concerned object from Figure-1. Figure-2.a shows region ‘a’ where the first guess of boundary from the prior knowledge is denoted by ‘a1’ and ‘a2’. Then we fine-tune this by the divide and conquer method. Each time we discover a point on the boundary by the perpendicular line from the midpoint of the guessed boundary line (here initial boundary is the a1a2 line Figure-2.a). Similar approach is followed for portion ‘b’ (Figure-2.b). The only difference is the search direction; one is outwards (Figure-2.a) other is inwards (Figure-2.b). Figure-1: Equal spaced parallel lines to determine points of initial guess on the object boundary o b a