Software complexity measurement is an essential part in the domain of software engineering. In measurement metrics developing, metrics was classed as two kinds, macro and micro. Macro complexity metrics consider the difficulty of system. Micro metrics are based on program code measuring. In order to enhance the measuring precision, people invent new metrics to replace famous metrics. But, many factors which cause imprecise measuring exist in measuring target. With normalization, the purpose of this paper is to improve the outcome consistency of computations of different metric complexities. Furthermore, the advantage of the paper is that normalized programs could derive complexity value of Halstead and Cyclomatic from a complexity value of LOC. We focus on micro measuring and structure program. We find out the imprecision factors which hide in syntax via studying C language. And, we use program normalization to improve accuracy of measuring.