Volume 3, Issue 11, November 2013 ISSN: 2277 128X

International Journal of Advanced Research in

Computer Science and Software Engineering
Research Paper
Available online at: www.ijarcsse.com
An Efficient Fence Fill Algorithm using Inside-Outside Test
S.Anitha, D.Evangeline
Department of Information Technology,
School of Communication and Computer Sciences,
Kongu Engineering College, Perundurai, Erode-638052,
Tamil Nadu, India

Abstract— In this paper, we introduce a modified fence fill algorithm for filling a polygon. Fence fill algorithm is a
modified edge fill algorithm. As opposed to the edge fill algorithm that fills pixels starting from the polygon edge to
the maximum screen coordinates, the fence fill algorithm fills pixels bounded by the polygon edge and the fence. This
involves large number of calculations to find the intersection of the polygon edge with the scan line. Our method
reduces the calculations by considering a single pixel in each region bounded by the fence and polygon edge. The
inside-outside test is then applied to that pixel to determine whether the point lies inside or outside the bounded region.
For large concave polygons, the older method is complicated as it requires more space and time. The new method
promises a significant reduction in memory and time requirements.

Keywords— Polygon fill, Edge fill, Fence fill, Inside-outside test

I. INTRODUCTION
Computer Graphics are graphical image created using computers with specialized graphics hardware and software.
Images are composed of basic structures like points, lines, curves, polygons, etc., Knowledge of Graphics is essential in
the field of image processing which finds its applications in research, medicine, steganography, satellite communication,
remote sensing, biometrics and so on. A polygon is a closed contour. Polygons are classified as convex and concave
(non-convex). Convex polygons are those polygons for which the line joining any two points within the polygon also lies
completely within the polygon. In computer graphics, vector method and rotational method are two methods used to
determine whether the polygon is concave or convex. If a polygon is concave, these methods split such polygons into a
pair of convex polygons. Polygon filling is the process of coloring the pixels of a polygon. Many polygon filling
algorithms are employed to fill polygons in various graphics packages. Some of the traditional polygon fill algorithms
include Scan Line Polygon Fill, Seed Fill and Edge Fill. Such algorithms can be applied to both concave and convex
polygons but is generally more complicated for concave polygons. The scan line algorithm computes the intersection of
the polygon edge with each scan line. For each scan line crossing the polygon edge, the intersection coordinates are
sorted left to right and the pixels in each scan line are colored in a very similar manner. Boundary Fill and Flood Fill
algorithms fall under Seed fill since these algorithms commence with the assumption of a known interior pixel within the
polygon. Seed Fill algorithms that fill interior defined regions are known as Flood Fill and those seed fill algorithms that
fill boundary defined regions are known as Boundary Fill. Both the seed fill algorithms may be 4-connected or 8-
connected. Edge Fill algorithm fills pixels from the intersection coordinate of polygon edge with the scan line to the
extreme screen coordinates for each scan line. Hence, the pixels lying exterior to the polygon will be colored twice
resulting in the initial unmarked condition. Fence Fill is a modified edge fill algorithm wherein a fence is constructed
such that it passes through any vertex of the polygon. The pixels lying in the region bounded by the fence and the
polygon edge is colored. All these algorithms work in raster coordinate systems which may be region-oriented or vector-
oriented.
Considering polygon filling as a problem of coloring interior pixels of a polygon, we propose a modified and efficient
fence fill algorithm.
(1) Construct a fence that passes through the vertex or center of a polygon edge;
(2) Apply inside-outside test to a single pixel lying in each region created by the polygon edge with respect to the fence;
(3) If the pixel of the region lies within a polygon, color all the pixels lying in the region.
The remainder of the paper is organized as follows. In section I, an overview of the conventional polygon splitting,
inside-outside tests and filling algorithms required is discussed. In section II, the proposed algorithm is illustrated. We
draw our conclusion and results in the last section.
II. BACKGROUND
A. Test to determine whether a polygon is concave or convex[2]
1. Vector Method
For a convex polygon, the cross product of adjacent edges (ie., the z-component) holds the same sign. But, for a concave
polygon, the z-component turns out to be negative and the polygon is split along the line of the first edge vector in the
cross product pair.

© 2013, IJARCSSE All Rights Reserved Page | 605

 Anitha et al., International Journal of Advanced Research in Computer Science and Software Engineering 3(11),
November - 2013, pp. 605-609
2. Rotational Method
Process the polygon vertices in anti-clockwise direction. Translate the vertices of a polygon Vi to the coordinate origin.
Rotate the polygon in clockwise direction such that the next vertex Vi+1 lies on the x-axis. If the vertex Vi+2 lies below,
the polygon is concave and it is split along x-axis. If not, continue to rotate polygon vertices on x-axis and test for
negative y values.
A C
B
A B

E C

F D

Figure-1. Concave Polygon Figure-2. Convex Polygon

Using the vector and rotational method, the polygon ABC is determined as convex and the polygon ABCDEF is
determined as concave and the polygon is split along the edge AB.

B. Inside Outside Tests

1. Odd Even Rule
If the number of polygon edges crossed by the line starting at a position P to a distant point outside the object, then
P is an exterior point. Otherwise, P is an interior point.

2. Winding Number Rule

If the number of times the edges wind around any point P in anti-clockwise direction is non-zero, then P is an
interior point. Otherwise, P is an exterior point.

When the region bounded by EA, AB, BC, DE and CD is considered exterior, Odd Even rule and Winding Number
rule reveal that:
i. PP1 and PP4 are exterior
ii. PP2 and PP3 are interior

It is not necessary that both the rules may yield the same result.

P4

P P3

C P
P1

P2
E P B

Figure-3 Inside –Outside Tests

C. Polygon Filling Algorithms
The simplest polygon filling algorithm considers every pixel in the raster, verifies whether it lies within the polygon
and colors the pixel if it is interior. This technique is tedious since each and every pixel in the raster is considered.
Alternately, a bounding rectangle for the polygon ie., the smallest rectangle containing the polygon is constructed and the
pixels within the bounding rectangle is considered.

i. Scan Line Polygon Fill Algorithm

A polygon is scan converted by finding the intersection of all polygon edge on each scan line. The pixels within the
polygon are then determined and colored.

© 2013, IJARCSSE All Rights Reserved Page | 606

 Anitha et al., International Journal of Advanced Research in Computer Science and Software Engineering 3(11),
November - 2013, pp. 605-609

y- axis
A C

Scan line 8

F D

2 4 6 8 x-axis

Figure-4. Scan Line Polygon Fill

The intersection of polygon edge AF, AB, BC and CD on scan line 8 is found. The pixels on scan line 8 spanning from
x=2 to x=4 and x=7 to x=8 are colored.

ii. Seed Fill Algorithm

The seed fill algorithms (boundary fill and flood fill) are highly recursive. These algorithms may be 8-connected or 4-
connected.

○ × ○
× × ×
× ○ ×
× ○ ×
○ × ○
× × ×

4-connected 8-connected

Figure-5. Seed Fill

Seed Fill algorithms accept as input: a known interior pixel (seed pixel) and fill color. Boundary Fill algorithms
start with the seed pixel, check whether their neighbors (4-connected or 8-connected) are of boundary color and if not,
they are filled with fill color. This is recursively done until pixel with boundary color is encountered. When the interior
region is marked with multi-colored or no boundary, flood fill algorithm is adopted. Here, the interior pixels are filled
with a specific color. Check neighbors (4-connected or 8-connected) if they are of interior color. If so, fill such pixels
with fill color.

iii. Edge Fill algorithm

The algorithm finds the intersection of the polygon edge with each scan line, colors pixels from the intersection to
the extreme screen coordinates for each scan line. This is done for all polygon edges. Interior pixels within the polygon
are colored once and exterior pixels are colored twice.

This algorithm can be restated as:―For each scan line intersecting the polygon edge at (x 1,y1), complement all the
pixels whose midpoints lie to the right of (x1,y1) i.e., for (x,y1), x+1/2>x1.‖[3]

iv. Fence Fill algorithm

This algorithm also finds the intersection of the polygon edge with each scan line. However, coloring is done based
on the position of the pixels with respect to the fence (a line that passes through any vertex or polygon edge). Each pixel
in the region created by the polygon edge and fence is checked whether it is interior or exterior.

III. PROPOSED ALGORITHM

The algorithm focuses on filling concave polygons.A fence location is usually considered passing through any polygon
vertex for simplicity. A bounding rectangle for the polygon is constructed. The regions bounded by the polygon edge
with the fence are identified. A single pixel in each region is chosen and inside-outside tests are carried on it. If the test
returns interior, the pixel is colored. Otherwise, the pixel remains uncolored.

© 2013, IJARCSSE All Rights Reserved Page | 607

 Anitha et al., International Journal of Advanced Research in Computer Science and Software Engineering 3(11),
November - 2013, pp. 605-609
The algorithm is illustrated in the following figures. II III

A C
A C B
B I IV

IX
V

E VIII E
F D
F D

VII VI

Figure-6 Fence and bounding rectangle Figure-7 Identification of regions

Consider an imaginary line passing through the intersected point between the edge and the fence along the scan line.
For those edges not intersecting at the fence, extend the corresponding edge towards the fence. The different regions are
considered based on closed area formed between imaginary line, edge and fence. Inside-outside test is conducted on a
single pixel on each region. Since the regions I, IV, V, VII and IX are interior, the pixels in these regions are filled.
The proposed method shown in the Figure-7 which reduces the number of calculations compared to existing algorithms.
Consider a large screen display A(0,1000), B(1000,1000), C(1000,2000) and D(0,2000).
Convex Polygon
Case 1: Three sided A(100,100) B(100,200) C(150,150).
Case 2: Four sided A(50,150) B(100,100) C(100,200) D(150,150).
Case 3: Five sided A(50,150) B(100,220) C(150,180) D(120,100) and E(100,50).

(a) (b) (c)

Fig 8 Convex Polygon (a) Case 1 (b) Case 2 (c) Case 3

Concave Polygon
Case 1: Four sided A(125,150) B(100,120) C(125,200) D(150,120).
Case 2: Five sided A(100,120) B(85,205) C(120,180) D(165,215) and E(150,150).
Case 3: Six sided A(50,100) B(50,150) C(90,140) D(125,155) , E(100,120) and F(75,125).

(a) (b) (c)

Fig 9 Concave Polygon (a) Case 1 (b) Case 2 (c) Case 3

IV. PERFORMANCE EVALUATION

In Table 1, a comparison on Edge Fill, Fence Fill and the proposed algorithm is made based on the number of
calculations done to find the intermediate pixels on the polygon’s edge intersecting the scan line and the number of
comparisons required to determine the position of the pixel (interior or exterior to the polygon). For comparison, three
cases in both concave and convex polygons are considered. The edge fill and fence fill algorithms require calculations for
determination of intermediate pixels. In our proposed algorithm does not require any such calculations. In all the three

© 2013, IJARCSSE All Rights Reserved Page | 608

 Anitha et al., International Journal of Advanced Research in Computer Science and Software Engineering 3(11),
November - 2013, pp. 605-609
algorithms, the position of the pixels must be identified. The edge fill algorithm requires large number of computations.
In cases where a small polygon is displayed on a large screen area, the number of calculations is very high. The fence fill
algorithm also requires such calculations but the number of calculations is considerably very much lesser than edge fill
algorithm. This is due to the construction of the fence at the most appropriate location. Our proposed algorithm also
requires such calculations but it selects only one pixel per region and determines the position of the pixel. Thereby, the
number of comparisons made is negligible.

Table 1 Performance Evaluation of various Polygon Filling Algorithms

Algorithms
Type of Edge Fill Fence Fill Proposed
Case
polygo Fence
s No of No of No of No of No of
n Location No of
intersection comparison intersection comparison intersection
(Vertical comparisons
calculations s calculations s calculations
Line)
1 200 260100 200 7550 150 0 3
Convex 2 200 262650 200 5100 100 0 8
3 340 171345 340 4115 100 0 8
1 250 69679 250 5000 125 0 8
Concav
e 2 240 22377 240 6815 120 0 4
3 140 133800 140 1750 140 0 9

V. CONCLUSION
Our algorithm considerably reduces number of comparisons for polygon filling using Edge Fill and Fence Fill
algorithms. In traditional Edge Fill algorithm, not only the pixels interior to the polygon but also the pixels exterior to the
polygon are considered for comparison to fill the polygon. This increases both the memory requirements and execution
time. In cases where a small polygon is displayed on large screen, the number of pixels considered for comparison
greatly increases. But in Fence Fill, appropriate selection of fence drastically reduces such comparisons. Yet, the number
of comparisons is more. Our algorithm focuses only the polygonal area for comparison to fill it. Irrespective of the size of
the screen, the number of comparisons made is proportional to the size of the polygon (ie., the number of pixels interior
to the polygon). Moreover, not all the pixels interior to the polygon are considered for comparison. Only a pixel per
region is considered for comparison in our proposed algorithm. However, the accuracy of the algorithm totally depends
on the results of the Inside-Outside Tests.

REFERENCES
[1] Michael R Dunlavey. Efficient Polygon Filling Algorithms for Raster Displays. ACM Transactions on Graphics,
Vol.2, No 4, October 1983, Pages 264-273.
[2] Donald D.Hearn, M.Pauline Baker, Computer Graphics C Version, Second Edition, Pearson Education.
[3] David F. Rogers, Procedural Elements of Computer Graphics, First Edition, McGraw Hill
[4] Foley van Dam, Feiner Hughes, Computer Graphics Principles & Practice, Second Edition in C, Pearson
Education.
[5] Steven Harrington,Computer Graphics A Programming Approach, Second Edition, McGraw Hill
[6] A New Polygon Based Algorithm for Filling Regions, Hoo-Cheng Liu, Mu-Hwa Chen, Shou-Yiing Hsu, Chaoyin
Chien, Tsu-Feng Kuo and Yih-Farn Wang, Tamkang Journal of Science and Engineering, Vol. 2, No. 4, pp. 175-
186 (2000).

© 2013, IJARCSSE All Rights Reserved Page | 609