Let's say i have the following:

  • Point somewhere in space
  • Camera with position and orientation (up, right, forward)

I want to rotate camera around the point, but also keep this point in same place on screen. So, if point was on (32, 32) on window, after rotation i want it to still be on (32, 32).

I've seen How can I orbit a camera about it's target point? , and it was somewhat helpful. I needed code to rotate point around arbitrary axis (camera's up and right), so i used this resource.

Problem is, i got something like numerical errors and my camera started to wander weirdly when rotating around camera's both up and right (it seems fine when i rotate only around one of them).

I tested my implementation with code:

Matrix m1=MatrixRotate(Vector(-1, -1, 1),  33);
Matrix m2=MatrixRotate(Vector(-1, -1, 1), -33);
Vector a=Vector(-1, -1, -1);
Vector c=a;

printf("%f %f %f %f\n", c.x, c.y, c.z, c.w);

And got:

-1.028036 -0.960396 -1.124331 0.000000

It worked fine when rotation axis was something 'normal' like (1, 0, 0) or (0, 0, -1).

So, how else can i rotate camera around point, while keeping said point in same point on screen?


Looks like you got some very big rounding errors in there. I have no idea why but I have written some code for matrix rotations once. It's in Java though but should be very easy to translate to C++.

Vector class:

public class Vec {
    public final static Vec X_AXIS=new Vec(1,0,0);
    public final static Vec Y_AXIS=new Vec(0,1,0);
    public final static Vec Z_AXIS=new Vec(0,0,1);
    public final static Vec ORIGIN=new Vec(0,0,0);

    public final double x,y,z;

    public Vec(double x, double y, double z) {

    public double dotProduct(Vec v) {
        return x*v.x+y*v.y+z*v.z;

    public Vec crossProduct(Vec v) {
        double rx=y*v.z-z*v.y;
        double ry=z*v.x-x*v.z;
        double rz=x*v.y-y*v.x;
        return new Vec(rx, ry, rz);

    public Vec newLength(double newLength) {
        double length=getLength();
        if(length==newLength) return this;
        if(length==0) return X_AXIS.newLength(newLength);
        return new Vec(x*newLength/length, y*newLength/length, z*newLength/length);

    public Vec rotationAxis() {
        return rotationAxis(X_AXIS);

    //The rotation axis to rotate v onto this
    public Vec rotationAxis(Vec v) {
        return normalizedCrossProduct(v);

    public Vec normalizedCrossProduct(Vec v) {
        Vec r=crossProduct(v);
        if(r.getLength()<0.0001) {
        if(r.getLength()<0.0001) {
        if(r.getLength()<0.0001) {
            return X_AXIS;
        return r.normalize();

    public double angle() {
        return angleTo(X_AXIS);

    public double angleTo(Vec v) {
        double cosTheta=dotProduct(v)/(getLength()*v.getLength());

        return Math.acos(cosTheta);

    public double getLength() {
        return Math.sqrt(x*x+y*y+z*z);

    public Vec rotate(Matrix m) {
        return m.mul(this);

    public Vec normalize() {
        double length=getLength();

        if(length==1) return this;
        if(length==0) return X_AXIS;
        return new Vec(x/length, y/length, z/length);

    public Vec addX(double a) {
        return new Vec(x+a, y, z);

    public Vec addY(double a) {
        return new Vec(x, y+a, z);

    public Vec addZ(double a) {
        return new Vec(x, y, z+a);

    public Vec add(Vec v) {
        return new Vec(x+v.x, y+v.y, z+v.z);

    public Vec sub(Vec v) {
        return new Vec(x-v.x, y-v.y, z-v.z);

    public Vec mul(double m) {
        return new Vec(x*m, y*m, z*m);

    public Vec div(double d) {
        return new Vec(x/d, y/d, z/d);

    public Vec neg() {
        return new Vec(-x, -y, -z);

    public boolean equals(Object o) {
        if(!(o instanceof Vec)) return false;
        Vec v=(Vec)o;
        return v==this || sub(v).getLength() < 0.0001;

    public String toString() {
        return String.format(java.util.Locale.ENGLISH,
            "(%.3f, %.3f, %.3f)", x, y, z);


Matrix class:

public class Matrix {
    public final static Matrix IDENTITY=new Matrix(Vec.X_AXIS, Vec.Y_AXIS, Vec.Z_AXIS);
    public final Vec xAxis, yAxis, zAxis;

    public Matrix(double ... e) {
        if(e.length!=9) throw new RuntimeException();
        xAxis = new Vec(e[0], e[3], e[6]);
        yAxis = new Vec(e[1], e[4], e[7]);
        zAxis = new Vec(e[2], e[5], e[8]);

    public Matrix(Vec xAxis, Vec yAxis, Vec zAxis) {
        this.xAxis = xAxis;
        this.yAxis = yAxis;
        this.zAxis = zAxis;

    public Matrix(Vec axis, double theta) {
        Vec u=axis.normalize();
        double sin=Math.sin(theta);
        double cos=Math.cos(theta);
        double uxy=u.x*u.y*(1-cos);
        double uyz=u.y*u.z*(1-cos);
        double uxz=u.x*u.z*(1-cos);
        double ux2=u.x*u.x*(1-cos);
        double uy2=u.y*u.y*(1-cos);
        double uz2=u.z*u.z*(1-cos);
        double uxsin=u.x*sin;
        double uysin=u.y*sin;
        double uzsin=u.z*sin;

        xAxis = new Vec(cos+ux2, uxy+uzsin, uxz-uysin);
        yAxis = new Vec(uxy-uzsin, cos+uy2, uyz+uxsin);
        zAxis = new Vec(uxz+uysin, uyz-uxsin, cos+uz2);

    static public Matrix xRotationMatrix(double theta) {
        double cos = Math.cos(theta), sin = Math.sin(theta);
        return new Matrix(Vec.X_AXIS,
                          new Vec(0, cos, sin),
                          new Vec(0, -sin, cos));

    static public Matrix yRotationMatrix(double theta) {
        double cos = Math.cos(theta), sin = Math.sin(theta);
        return new Matrix(new Vec(cos, 0, -sin),
                          new Vec(sin, 0, cos));

    static public Matrix zRotationMatrix(double theta) {
        double cos = Math.cos(theta), sin = Math.sin(theta);
        return new Matrix(new Vec(cos, sin, 0),
                          new Vec(-sin, cos, 0),

    public Matrix rotX(double theta) {
        return xRotationMatrix(theta).mul(this);

    public Matrix rotY(double theta) {
        return yRotationMatrix(theta).mul(this);

    public Matrix rotZ(double theta) {
        return zRotationMatrix(theta).mul(this);

    public Matrix mul(double d) {
        return new Matrix(xAxis.mul(d), yAxis.mul(d), zAxis.mul(d));

    public Vec mul(Vec v) {
        return xAxis.mul(v.x).add(yAxis.mul(v.y)).add(zAxis.mul(v.z));

    public Matrix mul(Matrix m) {
        return new Matrix(mul(m.xAxis), mul(m.yAxis), mul(m.zAxis));

    public Matrix rotateRel(Matrix m) {
        return mul(m);

    public Matrix rotateAbs(Matrix m) {
        return m.mul(this);

    public Matrix normalize() {
        Vec vz=xAxis.crossProduct(yAxis);
        Vec vy=vz.crossProduct(xAxis);        
        return new Matrix(xAxis.normalize(),

    public Matrix transpose() {
        return new Matrix(xAxis.x, xAxis.y, xAxis.z,
                          yAxis.x, yAxis.y, yAxis.z,
                          zAxis.x, zAxis.y, zAxis.z);

    public double determinant() {
        return xAxis.x*(yAxis.y*zAxis.z-zAxis.y*yAxis.z) -
               yAxis.x*(zAxis.z*xAxis.y-zAxis.y*xAxis.z) +

    public Matrix inverse() {
        Vec A = yAxis.crossProduct(zAxis);
        Vec B = zAxis.crossProduct(xAxis);
        Vec C = xAxis.crossProduct(yAxis);
        return new Matrix(A,B,C).transpose().mul(1/determinant());

    public Matrix oppositeRotMatrix() {
        return transpose();

    public boolean equals(Object o) {
        if(!(o instanceof Matrix)) return false;
        Matrix m=(Matrix)o;
        return m==this || xAxis.equals(m.xAxis) && yAxis.equals(m.yAxis) && zAxis.equals(m.zAxis);

    public String toString() {
        return String.format(java.util.Locale.ENGLISH,
            "[%.3f, %.3f, %.3f]\n[%.3f, %.3f, %.3f]\n[%.3f, %.3f, %.3f]",
            xAxis.x, yAxis.x, zAxis.x,
            xAxis.y, yAxis.y, zAxis.y,
            xAxis.z, yAxis.z, zAxis.z);

When I test it with this code

Matrix m1 = new Matrix(new Vec(-1, -1, 1), Math.toRadians(33));
Matrix m2 = new Matrix(new Vec(-1, -1, 1), Math.toRadians(-33));
Vec a = new Vec(-1, -1, -1);
Vec c = a;

c = m1.mul(c);
c = m2.mul(c);

I get (-1.000, -1.000, -1.000), so no significant rounding errors.

btw. I took my equations for axis/angle rotations from http://en.wikipedia.org/wiki/Rotation_matrix

  • 1
    \$\begingroup\$ May His Noodly Appendage touch you, my friend. Your code helped me get rid of numerical errors, but camera kept wandering when rotating. Solution: i changed order of rotations. Before, i rotated first by Up then by Right, now it is Right then Up. Thanks. \$\endgroup\$ – crueltear Apr 28 '14 at 11:40
  • \$\begingroup\$ You can multiply two matrices. For example matrix m1 rotates up and matrix m2 rotates right. Then you can do m1*m2 to get a matrix that rotates up and right. However with matrix multiplications the order matters. e.g. m1 * m2 would mean first rotate right (m2) and then rotate up (m1), where both rotations are in the absolute coordinate system. But m1*m2 could also mean, first rotate up and then rotate right in the relative coordinate system of the camera. Both points of view are mathematically equivalent. \$\endgroup\$ – SpiderPig Apr 28 '14 at 23:00

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.