Computing Euler angles from a rotation matrix

Gregory G. Slabaugh
Abstract
This document discusses a simple technique to find all possible Euler angles from
a rotation matrix. Determination of Euler angles is sometimes a necessary step
in computer graphics, vision, robotics, and kinematics. However, the solution
may or may not be obvious.

Rotation matrices
We start off with the standard definition of the rotations about the three principle axes.
A rotation of radians about the x-axis is defined as

1
0
0
Rx () = 0 cos sin
0 sin
cos
Similarly, a rotation of radians about

cos
0
Ry () =
sin

the y-axis is defined as

0 sin
1
0
0 cos

Finally, a rotation of radians about the z-axis is defined as

cos sin 0
Rz () = sin cos 0
0
0
1
The angles , , and are the Euler angles.

Generalized rotation matrices

A general rotation matrix can will have the form,

R11 R12 R13

R = R21 R22 R23
R31 R32 R33

This matrix can be thought of a sequence of three rotations, one about each
principle axis. Since matrix multiplication does not commute, the order of the
axes which one rotates about will affect the result. For this analysis, we will
rotate first about the x-axis, then the y-axis, and finally the z-axis. Such a
sequence of rotations can be represented as the matrix product,

= Rz ()Ry ()Rx ()

cos cos sin sin cos cos sin cos sin cos + sin sin
= cos sin sin sin sin + cos cos cos sin sin sin cos
sin
sin cos
cos cos

Given a rotation matrix R, we can compute the Euler angles, , , and

by equating each element in R with the corresponding element in the matrix
product Rz ()Ry ()Rx (). This results in nine equations that can be used to
find the Euler angles.

Finding two possible angles for

Starting with R31 , we find
R31 = sin .
This equation can be inverted to yield
= sin1 (R31 ).

(1)

However, one must be careful in interpreting this equation. Since sin( ) =

sin(), there are actually two distinct values (for R31 6= 1) of that satisfy
Equation 1. Therefore, both the values
1
2

= sin1 (R31 )
= 1 = + sin1 (R31 )

are valid solutions. We will handle the special case of R31 = 1 later in this
report. So using the R31 element of the rotation matrix, we are able to determine
two possible values for .

Finding the corresponding angles of

To find the values for , we observe that
R32
= tan().
R33
We use this equation to solve for , as
= atan2(R32 , R33 ),

(2)

where atan2(y, x) is arc tangent of the two variables x and y. It is similar to

calculating the arc tangent of y/x, except that the signs of both arguments are
used to determine the quadrant of the result, which lies in the range [, ].
The function atan2 is available in many programming languages.
One must be careful in interpreting Equation 2. If cos() > 0, then =
atan2(R32 , R33 ). However, when cos() < 0, = atan2(R32 , R33 ). A simple
way to handle this is to use the equation


R32 R33
,
(3)
= atan2
cos cos
to compute .
Equation 3 is valid for all cases except when cos = 0. We will deal with this
special case later in this report. For each value of , we compute a corresponding
value of using Equation 3, yielding


R32
R33
1 = atan2
,
(4)
cos 1 cos 1


R33
R32
(5)
2 = atan2
,
cos 2 cos 2

Finding the corresponding angles of

A similar analysis holds for finding . We observe that
R21
= tan .
R11
We solve for using the equation

= atan2

R21 R11
,
cos cos


(6)

Again, this equation is valid for all cases except when cos = 0. We will
deal with this special case later in this report. For each value of , we compute
a corresponding value of using Equation 6,


R21
R11
1 = atan2
,
(7)
cos 1 cos 1


R21
R11
2 = atan2
,
(8)
cos 2 cos 2

Two solutions if cos 6= 0

For the case of cos 6= 0, we now have two triplets of Euler angles that reproduce
the rotation matrix, namely
(1 , 1 , 1 )
(2 , 2 , 2 )
Both of these solutions will be valid.
3

What if cos = 0?
This technique described above does not work if the R31 element of the rotation
matrix is 1 or 1, which corresponds to = /2 or = /2, respectively, and
to cos = 0. When we try to solve for the possible values of and using
the above technique, problems will occur, since the elements R11 , R21 , R32 , and
R33 will all be zero, and therefore Equations 3 and 6 will become


0 0
,
= atan2
0 0


0 0
= atan2
,
.
0 0
In this case R11 , R21 , R32 , and R33 do not constrain the values of and .
Therefore, we must use different elements of the rotation matrix to compute the
values of and .

= /2 case: Consider the case when = /2. Then,

R12
R13
R22
R23

= sin cos cos sin = sin( )

= cos cos + sin sin = cos( )
= sin sin + cos cos = cos( ) = R13
= cos sin sin cos = sin( ) = R12

Any and that satisfy these equations will be a valid solution. Using
the equations for R12 and R13 , we find that
( ) = atan2(R12 , R13 )
= + atan2(R12 , R13 )

= /2 case: Not surprisingly, a similar result holds for the case when =
/2, for which
R12
R13
R22
R23

=
=
=
=

sin cos cos sin = sin( + )

cos cos + sin sin = cos( + )
sin sin + cos cos = cos( + ) = R13
cos sin sin cos = sin( + ) = R12

Again, using the equations for R12 and R13 , we find that
( + ) = atan2(R12 , R13 )
= + atan2(R12 , R13 )

if (R31 6= 1)
1 = asin(R31 )
2 = 1 

R32
R33
1 = atan2 cos
,
 1 cos 1 
R32
2 = atan2 cos
, R33
 2 cos 2 
R21
1 = atan2 cos
, R11
 1 cos 1 
R11
R21
2 = atan2 cos
2 , cos 2
else
= anything; can set to 0
if (R31 = 1)
= /2
= + atan2(R12 , R13 )
else
= /2
= + atan2(R12 , R13 )
end if
end if
Figure 1: Pseudo-code for computing Euler angles from a rotation matrix. See
text for details.
Either case: In both the = /2 and = /2 cases, we have found that
and are linked. This phenomenon is called Gimbal lock. Although
in this case, there are an infinite number of solutions to the problem, in
practice, one is often interested in finding one solution. For this task, it is
convenient to set = 0 and compute as described above.

Pseudo-code
We now summarize the method by providing a pseudo-code implementation in
Figure 1. The code is very simple.

Example
An example that demonstrates the computation of , , and from a rotation
matrix is provided below.
Suppose we are asked to find the Euler angles that produce the matrix

.5
.1464 .8536
.5
.8536 .1464
R=
.7071
.5
.5
First, we find the possible values for to be
1

= sin(.7071) =
5

= 1 =

3
4

Then, we find the corresponding values of to be



.5
.5

1 = atan2
,
=
cos(/4) cos(/4)
4


.5
.5
3
2 = atan2
,
=
cos(3/4) cos(3/4)
4
And we find to be


1
2


.5
.5

= atan2
,
=
cos(/4) cos(/4)
4


.5
.5
3
= atan2
,
=
cos(3/4) cos(3/4)
4

Therefore, the solutions are

 
, ,
4 4 4

3 3 3
,
,
4 4
4

More than one solution?

It is interesting to note that there is always more than one sequence of rotations
about the three principle axes that results in the same orientation of an object.
As we have shown in this report, in the non-degenerate case of cos 6= 0, there
are two solutions. For the degenerate case of cos = 0, an infinite number of
solutions exist.
As an example, consider a book laying on a table face up in front of you.
Define the x-axis as to the right, the y-axis as away from you, and the z-axis
up. A rotation of radians about the y-axis will turn the book so that the back
cover is now facing up. Another way to achieve the same orientation would be
to rotate the book radians about the x-axis, and then radians about the
z-axis. Thus, there is more than one way to achieve a desired rotation.

Acknowledgement
I wish to thank Sinisa Segvic, from the University of Zagreb, for providing some
insights into how to reduce the number of possible solutions to two for the
non-degenerate case. Also, many thanks to Tobias Ziegler from the Fraunhofer
Institute for pointing out the need for negative signs on R12 and R13 in the
= /2 case.

References
[1] Ken Shoemake, Animating Rotation with Quaternion Curves Proc. SIGGRAPH 1985, pp. 245 - 254.