AN EFFICIENT LINE CLIPPING ALGORITHM FOR CIRCULAR WINDOWS USING VECTOR CALCULUS AND PARALLELIZATION

International Journal of Computer Graphics & Animation (IJCGA) Vol.8, No.1/2, April 2018
DOI : 10.5121/ijcga.2018.8201 1
AN EFFICIENT LINE CLIPPING ALGORITHM FOR
CIRCULAR WINDOWS USING VECTOR CALCULUS AND
PARALLELIZATION
Prastut Kumar1
, Fenil Patel1
and Rajesh Kanna2
1&2
Department of Computing Science & Engineering, Vellore Institute of Technology,
Chennai Campus
ABSTRACT
With the advent of digitization and growing abundance of graphic and image processing tools, use cases
for clipping using circular windows have grown considerably. This paper presents an efficient clipping
algorithm for line segments using geometrical features of circle and vector calculus. Building upon the
research with rectangular windows, this method is proposed with the belief that computations are more
expensive (heavy) than other computations. Execution time can be drastically improved if we replace
expensive computations with cheaper computations. The cheaper computations can be computed even more
efficiently using parallelization thus improving time complexity.
KEYWORDS
Line clipping, Circle boundary, vector calculus, parallelization
1. INTRODUCTION
In computer graphics, line clipping algorithms with respect to rectangular windows has seen its
fair share of success and has become a basic operation in most graphic applications. Popular
algorithms like Cohen-Sutherland [1], Liang-Barsky[2], midpoint division [3] algorithm suited
for implementation purposes, Nicholl–Lee–Nicholl algorithm [4], FLC algorithm [5] have been
used extensively in production. Most algorithms come into two types, namely the encoding
approach and parametric approach.
Today in most graphic and image processing applications, the clipping window is no longer
restricted to a simple regular rectangle. This paper describes an algorithm based on circular
windows. In recent times, many line clipping algorithms for circular windows have been
presented. The two papers [6,7] are substituting the line segment parametric equation into the
algebraic equation of the circular window which is a very expensive computation in itself and an
additional drawback of ignoring the case that the line does not intersect the circle. Cai et al. [8]
uses position relation between the circular window and its outer tangent square, ruling out out the
above edge case, and then clips the line segment using the standardized intersection tables. The
lookup in searching tables is again expensive. The idea of Lu et al. [9] is applying multiple
encoding techniques.

2
Due to the advent of cheaply available GPU’s, parallel processing has been well accepted by the
scientific society, with many algorithms ported to support parallelization methods. This paper
proposes a clipping algorithm for line segments against circular windows using geometrical
features of circle and vector calculus. The method combines both encoding and parametric
approaches as we use the parametric equation of line segments for vector calculus and the
tangential property of the circle to clip the line. Since the algorithm works in the same fashion for
both endpoints of a line segment, we parallelize this process to reduce the execution time.
2. PROPOSED METHODOLOGY
We illustrate our methodology by first discussing the primitives of circle and how we use them
inside our algorithm, followed by the explanation of how our algorithm works using the very
same primitives.
Figure 1. The clipping of line segments against circular window.
Assume that the center of the circular window is at the origin of the coordinate system, the
equation of the circle is:
and the parametric equation of the line segment PQ is:
where the endpoints are given by P(x0 , y0 ) and Q(x1 , y1 ) . [see Fig. 1]
2.1 Preprocessing using Liang-Barsky Algorithm
The tangential square acts as the the clipping window for Liang-Barsky Algorithm. The choice
for Liang-Barsky was made since it’s one of the most efficient line clipping algorithms in terms
of implementation as well. A point on the line segment is in the clip window if

3
which can be expressed in terms of 4 inequalities
Where
To bound the tangential circle to the circle, we take the diameter parallel to the x axis and use the
left endpoint as xmin and the right endpoint as xmax and similarly use the diameter parallel to y axis
to compute ymin and ymax . To clip the the line segment, we need to consider various
cases:
1. A line parallel to a clipping window edge has pi = 0 for that boundary.
2. If for that i, qi < 0 , then the line is completely outside and can be eliminated.
3. When pi < 0 , the line proceeds outside to inside the clip window, and pi > 0 , the line proceeds
inside to outside.
4. For nonzero gives the pk , u = qi / pi intersection point.
5. For each line, calculate u1 and u2 . For u1 , look at the boundaries for which pi < 0 . For u2 ,
look at the boundaries for which pi > 0 .
6. Take u1 = max {0, qi / pi} and u2 = min {1 , qi / pi} . If u1 > u2 , the line is outside and
therefore rejected.
After applying the above algorithm by putting x0 = P(x0) & y0 = P(y0), we will be able to compute
the new endpoint P′(x0 , y0 ) . The same process is repeated with Q′. Finally, the new endpoints
given after applying Lian-Barsky Algorithm are P′& Q′ . [see Fig. 2]

4
Fig. 2: Using the tangential square as clipping window, applying Lian-Barsky results in new line
segment P′Q′ .
2.2. Line Clipping using vector calculus
As mentioned before, the center of the circle is at origin. Let it be denoted by O . We connect the
endpoints P′& Q′ to the center resulting in line segments OP′& OQ′ [See Fig. 3] . The line
segment OP′& OQ′ intersects the circle at points I1 & I2 respectively. To find intersection point I1
we don’t solve line equation with circle’s because it’s a very expensive computation since circle’s
equation in itself is a second degree equation. Instead we traverse a distance equal to the radius of
the circle on OP′& OQ′ to find I1 & I2 . We use vector calculus for traversing the distance and
finding the intersection points.

5
Fig. 3: a) Finding intersection point on circle of line segments OP′& OQ′. b) Zoom over endpoint
P′. For brevity purposes, let us first concentrate on & . To find P′ I1 I1 , we traverse a distance
equal to radius r along line segment OP′ . Let:
be the normalized vector parallel to OP′ and pointing to P′ . Therefore:
On I1 draw a tangent to the circle which would intersect our line segment P′Q′ . The intersection
point thus created on line segment P′Q′ is P′′ . Thus this is the new endpoint and the new clipped
line becomes P′′Q′ . [See Fig.4]
Fig. 4: Tangent drawn from I1 intersecting OP′ at P′′.

6
In similar fashion, we find I2 and then find Q′′ on line segment OQ′. The new clipped line
becomes P′′Q′′ . This happens repeatedly until the endpoints we get lie on the circle. Thus the
original line PQ gets clipped to become P′Q′.
Fig. 5: End of algorithm
2.3. Parallelization
In our proposed algorithm, the new endpoints are calculated in isolation i.e they have no
correlation in calculation with each other. Therefore we can parallelize endpoint calculation using
multithreading. Only two threads would be required for computation which will be fed by starting
points P & Q. Each thread execution produces a new endpoint which is again fed and the process
is repeated until the endpoints lie on the circle.
2.4 Handling edge cases
(1) Both endpoints lie inside of the circle . This results in case where line clipping is not
required at all. In the beginning of our method, substituting both endpoints in the equation of the
circle, we will check if both the endpoints lie in the circle or not:
If the result of the above operation for both endpoints is ≤ 0 that means the line segment lies
completely inside the circle and the algorithm exits. If not then we check for rest of the edge
cases.
(2) One of the endpoint lies inside the circle . If either of the endpoint returns a result ≥ 0, we
apply our algorithm for that endpoint only.
(3) Both endpoints lie outside of the circle . If both endpoints return a result ≥ 0 then both the
end points lie outside of the circle. This results in 2 edge cases i.e if the line lies completely
outside the tangential square or the line lies outside of the circle but lies inside the tangential
square [see Fig. 6].

7
b)
Fig. 6: Showing both cases.
For the first edge case, the first step of our algorithm uses Liang Barsky where the tangential
square is the clipping window. Liang-Barsky will return false if the line lies completely outside
the tangential square.
For the second edge case, we draw a perpendicular from the center of the circle to the line. If the
distance between the point of intersection of line and perpendicular is ≥ r that means either the
line lies outside the circle or the line is a tangent. In both cases, we exit without clipping. Finally,
if the distance is < r the line lies inside the circle and we can begin clipping.
3. CONCLUSION AND DISCUSSION
The algorithm proposed produces the exact same results as with any other line clipping
algorithms against circular windows. This paper provides an efficient method to clip line
segments against circular windows using modern day parallel processors. The proposed method is
extremely fast for the usually occurring case of line segment which lies completely in the circle

8
having both endpoints outside the circle, taking only constant time for execution. Combined with
parallelization, the execution time becomes half of original.
In our literature survey, most proposed algorithms date back to the time when GPUs and parallel
processing concepts were starting to develop and focus was on optimizing algorithms to use
minimum resources. Now since these tools and concepts have become cost effective, our take is
to use them further increase computation speed of our algorithm.
Last, we believe this methodology, albeit a few changes, could be easily applied to clipping of
segments like parabola, hyperbola etc.
REFERENCES
[1] Newman WM, Sproull RF. Principles of interactive computer graphics. New York: McGraw-Hill;
1979.
[2] Liang YD, Barsky BA. A new concept and method for line clipping. ACM Transactions on Graphics
1984;3(1):1–22.
[3] Sproull RF, Sutherland IE. A clipping divider. In: Proceedings of Fall Joint Computer Conference.
Washington: Thompson Books; 1968. p. 765–75.
[4] Nicholl TM, Lee DT, Nicholl RA. An efficient new algorithm for 2-D line clipping: Its development
and analysis. Computer Graphics 1987;21(4):253–92.
[5] Wang J, Liang Y, Peng Q. A 2-D line clipping algorithm with the least arithmetic operations. Chinese
Journal of Computers 1991;14(7):495–504 (in Chinese).
[6] Yao H, Song P, Zhang G. Clipping algorithm and practice of circular window. Journal of Computer-
Aided Design & Computer Graphics 1992;4(3):14–20 (in Chinese).
[7] Liu Y. Circular and elliptic clipping windows. Computer Engineering and Design 1994;15(4):33–7
(in Chinese).
[8] Cai M, Yuan C, Song J, Cai S. A fast line clipping algorithm against circular windows. Journal of
Computer-Aided Design & Computer Graphics 2001;13(12):1063–7 (in Chinese).
[9] Lu G, Xing J, Tan J. New clipping algorithm for line against circular window with multi-encoding
approach. Journal of Computer-Aided Design & Computer Graphics 2002;12:1133–7 (in Chinese).

AN EFFICIENT LINE CLIPPING ALGORITHM FOR CIRCULAR WINDOWS USING VECTOR CALCULUS AND PARALLELIZATION

More Related Content

What's hot (17)

Similar to AN EFFICIENT LINE CLIPPING ALGORITHM FOR CIRCULAR WINDOWS USING VECTOR CALCULUS AND PARALLELIZATION (20)

Recently uploaded (20)

AN EFFICIENT LINE CLIPPING ALGORITHM FOR CIRCULAR WINDOWS USING VECTOR CALCULUS AND PARALLELIZATION