Mid-point Ellipse Algorithm

Midpoint Ellipse Algorithm

Midpoint ellipse algorithm is a method for drawing ellipses in computer graphics.

This method is modified from Bresenham’s algorithm. The advantage of this modified
method is that only addition operations are required in the program loops. This leads
to simple and fast implementation in all processors.

Let us consider one quarter of an ellipse. The curve is divided into two regions. In
region I, the slope on the curve is greater than –1 while in region II less than –1.

y
Region I
b
dy/dx = –1

Region II
a x
m = –1

Consider the general equation of an ellipse,

b2x2 + a2y2 – a2b2 = 0
where a is the horizontal radius and b is the vertical radius, we can define an function
f(x,y) by which the error due to a prediction coordinate (x,y) can be obtained. The
appropriate pixels can be selected according to the error so that the required ellipse is
formed. The error can be confined within half a pixel.

Set f(x,y) = b2x2 + a2y2 – a2b2

In region I (dy/dx > –1),

(xk, yk)

SE

Prediction
(xk+1, yk–½)
Region I

x is always incremented in each step, i.e. xk+1 = xk + 1.

yk+1 = yk if E is selected, or yk+1 = yk – 1 if SE is selected.
In order to make decision between S and SE, a prediction (xk+1, yk–½) is set at the
middle between the two candidate pixels. A prediction function Pk can be defined as
follows:

Page 1 of 5
 Mid-point Ellipse Algorithm

Pk = f(xk+1, yk–½)
= b2(xk+1)2 + a2(yk–½)2 – a2b2
= b2(xk2 + 2xk + 1) + a2(yk2 – yk + ¼) – a2b2

If Pk < 0, select E :

Pk+1E = f(xk+2, yk–½)

= b2(xk+2)2 + a2(yk–½)2 – a2b2
= b2(xk2 + 4xk + 4) + a2(yk2 – yk + ¼) – a2b2

Change of PkE is: ∆PkE = Pk+1E – Pk = b2(2xk + 3)

If Pk > 0, select SE :

Pk+1SE = f(xk+2, yk–3/2)

= b2(xk+2)2 + a2(yk–3/2)2 – a2b2
= b2(xk2 + 4xk + 4) + a2(yk2 – 3yk + 9/4) – a2b2
Change of PkSE is ∆PkSE = Pk+1SE – Pk = b2(2xk + 3) – 2a2(yk – 1)

Calculate the changes of ∆Pk:

If E is selected,
∆Pk+1E = b2(2xk + 5)
∆2PkE = ∆Pk+1E – ∆PkE = 2b2

∆Pk+1SE = b2(2xk + 5) – 2a2(yk – 1)

∆2PkSE = ∆Pk+1SE – ∆PkSE = 2b2

If SE is selected,
∆Pk+1E = b2(2xk + 5)
∆2PkE = ∆Pk+1E – ∆PkE = 2b2

∆Pk+1SE = b2(2xk + 5) – 2a2(yk – 2)

∆2PkSE = ∆Pk+1SE – ∆PkSE = 2(a2 + b2)

Initial values:
x0 = 0, y0 = b, P0 = b2 + ¼a2(1 – 4b)
∆P0E = 3b2, ∆P0SE = 3b2 – 2a2(b – 1)

In region II (dy/dx < –1), all calculations are similar to that in region I except that y is
decremented in each step.

Page 2 of 5
 Mid-point Ellipse Algorithm

(xk, yk)

SE

Prediction
(xk+½, yk–1)
Region II

y is always decremented in each step, i.e. yk+1 = yk – 1.

xk+1 = xk if S is selected, or xk+1 = xk + 1 if SE is selected.

Pk = f(xk+½, yk–1)
= b2(xk+½)2 + a2(yk–1)2 – a2b2
= b2(xk2 + xk + ¼) + a2(yk2 – 2yk + 1) – a2b2

If Pk > 0, select S :

Pk+1S = f(xk+½, yk–2)

= b2(xk+½)2 + a2(yk–2)2 – a2b2
= b2(xk2 + xk + ¼) + a2(yk2 – 4yk + 4) – a2b2

Change of PkS is: ∆PkS = Pk+1S – Pk = a2(3 – 2yk)

If Pk < 0, select SE :

Pk+1SE = f(xk+3/2, yk–2)

= b2(xk+3/2)2 + a2(yk–2)2 – a2b2
= b2(xk2 + 3xk + 9/4) + a2(yk2 – 4yk + 4) – a2b2
Change of PkSE is ∆PkSE = Pk+1SE – Pk = 2b2(xk + 1) + a2(3 – 2yk)

Calculate the changes of ∆Pk:

If S is selected,
∆Pk+1S = a2(5 – 2yk)
∆2PkS = ∆Pk+1S – ∆PkS = 2a2

∆Pk+1SE = 2b2(xk + 1) + a2(5 – 2yk)

∆2PkSE = ∆Pk+1SE – ∆PkSE = 2a2

If SE is selected,
∆Pk+1S = a2(5 – 2yk)
∆2PkS = ∆Pk+1S – ∆PkS = 2a2

∆Pk+1SE = 2b2(2xk + 2) – a2(5 – 2yk)

Page 3 of 5
 Mid-point Ellipse Algorithm

∆2PkSE = ∆Pk+1SE – ∆PkSE = 2(a2 + b2)

Determine the boundary between region I and II:

dy − bx
Set f(x, y) = 0, = .
dx a 1 − x 2 a 2
2

a2 b2
When dy/dx = –1, x = and y = .
a 2 + b2 a 2 + b2
a2 b2
At region I, dy/dx > –1, x < and y > , therefore
a 2 + b2 a 2 + b2
 2a 2   b2 
∆PkSE < b2  + 3  − 2a 2  − 1 = 2a 2 + 3b2 .
 a 2 + b2   a +b
2 2


Initial values at region II:

a2 b2
x0 = and y 0 =
a 2 + b2 a 2 + b2
x0 and y0 will be the accumulative results from region I at the boundary. It is not
necessary to calculate them from values of a and b.
P0 = PkI – ¼[a2(4y0 – 3) + b2(4x0 + 3)] where PkI is the accumulative result from region
I at the boundary.
∆P0E = b2(2x0 + 3)
∆P0SE = 2a2 + 3b2

Implementation of the algorithm:

The algorithm described above shows how to obtain the pixel coordinates in the first
quarter only. The ellipse centre is assumed to be at the origin. In actual
implementation, the pixel coordinates in other quarters can be simply obtained by use
of the symmetric characteristics of an ellipse. For a pixel (x, y) in the first quarter, the
corresponding pixels in other three quarters are (x, –y), (–x, y) and (–x, –y)
respectively. If the centre is at (xC, yC), all calculated coordinate (x, y) should be
adjusted by adding the offset (xC, yC). For easy implementation, a function
PlotEllipse() is defined as follows:

PlotEllipse (xC, yC, x, y)

putpixel(xC+x, yC+y)
putpixel(xC+x, yC–y)
putpixel(xC–x, yC+y)
putpixel(xC–x, yC–y)
end PlotEllipse

The function to draw an ellipse is described in the following pseudo-codes:

DrawEllipse (xC, yC, a, b)

Declare integers x, y, P, ∆PE, ∆PS, ∆PSE, ∆2PE, ∆2PS and ∆2PSE
// Set initial values in region I

Page 4 of 5
 Mid-point Ellipse Algorithm

Set x = 0 and y = b
P = b2 + (a2(1 – 4b) – 2)/4 // Intentionally –2 to round off the value
∆PE = 3b2
∆2PE = 2b2
∆PSE = ∆PE – 2a2(b – 1)
∆2PSE = ∆2PE + 2a2
// Plot the pixels in region I
PlotEllipse(xC, yC, x, y)
while ∆PSE < 2a2 + 3b2
if P < 0 then // Select E
P = P + ∆PE
∆PE = ∆PE + ∆2PE
∆PSE = ∆PSE + ∆2PE
else // Select SE
P = P + ∆PSE
∆PE = ∆PE + ∆2PE
∆PSE = ∆PSE + ∆2PSE
decrement y
end if
increment x
PlotEllipse(xC, yC, x, y)
end while
// Set initial values in region II
P = P – (a2(4y – 3) + b2(4x + 3) + 2) / 4 // Intentionally +2 to round off the value
∆PS = a2(3 – 2y)
∆PSE = 2b2 + 3a2
∆2PS = 2a2
// Plot the pixels in region II
while y > 0
if P > 0 then // Select S
P = P + ∆P E

∆PE = ∆PE + ∆2PS

∆PSE = ∆PSE + ∆2PS
else // Select SE
P = P + ∆PSE
∆PE = ∆PE + ∆2PS
∆PSE = ∆PSE + ∆2PSE
increment x
end if
decrement y
PlotEllipse(xC, yC, x, y)
end while
end DrawEllipse

Page 5 of 5