The symmetrical and asymmetrical distribution of noise correlation between digital image and digital camera are discovered in the experiment of identifing digital image origin by noise correlation. In order to improve the validity of identifi cation of digital image origin and predetermine accurately threshold of identification

a new probability density function of generalized chi square distribution is constructed within (0

1］ for asymmetrical distribution of noise correlation. A new method is proposed to apply generalized chi square distribution to describe the asymmetrical distributions of noise correlation

and using generalized Gaussian distribution to describe the symmetrical distributions of noise correlation. The experiment shows that the new method can reduce over 38% of the false rate compared with the method of generalized Gaussian distribution

consequently

it is an effective approach to reduce the false rate by diverse probability density functions to describe different distributions of noise correlation.