A new algorithm for distortion invariant optical correlation pattern recognition based on amplitude-modulated phase-only filtering and sliced orthogonal nonlinear generalized (SONG) correlation is proposed. An algorithm for distortion invariant pattern recognition is proposed by utilizing amplitude-modulated phase-only filtering with good discrimination and synthetic discriminant function, and the result of correlation output under the algorithm is much better than that of conventional synthetic discriminant function. The technique of SONG correlation with excellent discrimination and performance of resisting noise is introduced, and the new algorithm with several excellent capabilities for distortion invariant pattern recognition and combining these two techniques of nonlinear filtering is obtained. The results of simulation indicate that the new algorithm not only enhances the quality of correlation peak and multiple target discrimination, but also has a good performance of resisting noise. The discrimination capacity of more than 97% can be obtained under new algorithm even the input image is destroyed by noise badly. Compared with other algorithm for distortion invariant pattern recognition combined with SONG correlation, the new algorithm is more practical and generalized with better discrimination.