Looking upon the gray level histogram as a mixture of two Gaussian density functions is a conventional model in the image segmentation

unfortunately the histogram of the complex image often appears a multi-peak feature. In order to get a more accuracy approximation of this kind of histogram

this paper generalizes this model by considering the histogram a mixture of several Gaussian density functions

and employs a new algorithm of Adaptive Oriented Orthogonal Projective Decomposition to handle the mathematical problems involved in this process. In this proposed method

the key parameters of each Gaussian function can be calculated efficiently

which adequately leads to the determination of the optimal thresholds between different neighboring Gaussian functions. A new parameter called the Dividual Ratio of Threshold is introduced and used as the reference for the selection of the final thresholds. Experimental results show that this method can be effectively applied for the multi-threshold segmentation of complex images.