<p>The complexity of chaotic systems is closely related to the randomness of the pseudo-random sequences they generate. To generate highly random and low-correlation pseudo-random sequences, this paper proposes an improved De-Jong map based on locally active memristor. The model exhibits numerous fixed points, most of them are unstable and shift as the parameters change. Through numerical simulations, the complexity enhancement behavior is verified under different parameters and initial conditions, including the expansion of the chaotic region and an increase in chaotic complexity. Coexisting attractors induced by initial conditions are found in bifurcation diagrams. A performance comparison between improved map and the original map confirms that the introduction of the locally active memristor significantly enhances system complexity, as shown by excellent dynamic performance metrics. Randomness tests also confirmed that the model can generate pseudo-random sequences with strong randomness. Additionally, a digital signal processing (DSP)-based hardware experimental platform is developed to validate the numerical results. Finally, a density-tunable pseudo-random sequence generation method based on this new map is applied to information hiding. The results show that this method can effectively improve bits per pixel and system robustness.</p>