Well-conditioned matrices are often required in science and engineering, such as signal processing and finance. Problem to find the nearest positive definite matrix by explicitly imposing a constraint on the condition number are considered in this paper. A new algorithm based on geometric perspective is proposed for getting the required well-conditioned matrix. Based on these, a condition number-constrained matrix minimization problem is further considered, where the constraints are imposed for avoiding degenerate solutions in which parameter matrices become rank deficient. An inexact version of alternating direction method with truly implementable inexactness criteria is proposed for solving this problem. Numerical experiments, including comparison with some existing methods, are performed to illustrate the efficiency of the proposed algorithms.