What is the definition of dimension, you are working with?
If as ``dimension" you consider the supremum of cardinalities of linearly independent subsets, then the answer is ``yes". The boundedness of operator is not significant.
It seems to me that with this definition the answer is also positive. Let xi be elements of orthonormal basis of K. Then, for every non-zero x in H there is an i such that is non-zero. This is the same as is non-zero, which means that T*xi form a complete system in H. The cardinality of a complete system must be greater than or equal to the cardinality of orthonormal basis in H.
I think orthogonal its is a red herring. By the Axiom of Choice, H has an (algebraic) basis that is transformed by T to an independent subset of K, which can be extended to a basis of K.