1: \begin{abstract}

2: Quaternion matrices are employed successfully in many color image processing applications.

3: In particular, a pure 

4: quaternion matrix can be used to represent red, green and blue channels of color images.

5: A low-rank approximation for a 

6: pure quaternion matrix can be obtained by using the quaternion singular value decomposition.

7: However, this approximation is not optimal in the sense that the resulting low-rank approximation matrix may not be pure quaternion, i.e.,

8: the low-rank matrix contains real component which is not useful for the representation of a color image.

9: The main contribution of this paper is to find an optimal rank-$r$ 

10: pure quaternion matrix approximation for a 

11: pure quaternion matrix (a color image).

12: Our idea is to use

13: a projection on a low-rank quaternion matrix manifold and a projection on a quaternion matrix with zero real component, and develop

14: an alternating projections algorithm to find such optimal low-rank 

15: pure quaternion matrix approximation.

16: The convergence of the projection algorithm

17: can be established by showing that the low-rank quaternion matrix manifold and the zero real component quaternion matrix manifold

18: has a non-trivial intersection point.

19: Numerical examples on synthetic 

20: pure quaternion matrices and color images

21: are presented to illustrate the projection algorithm can find optimal low-rank 

22: pure quaternion approximation

23: for pure quaternion matrices or color images.

24: \end{abstract}

25: