Denoising diffusion probabilistic models (DDPM) have had remarkable success in image generation. Inspired by this, recent medical image segmentation tasks have started to use diffusion-based methods. These methods leverage iterations and sampling to generate smoother and more representative implicit integration. However, current diffusion-based segmentation models mainly rely on traditional neural networks and seldom focus on effectively interacting semantic and noise information. Moreover, they usually use a single network architecture instead of a hybrid one combining CNN and Transformer. To address limitations, we propose a dual-path U-Net segmentation diffusion (DPUSegDiff) model. It comprises two U-shaped networks based on the edge augmented local encoder (EALE) and the mixed transformer global encoder (MTGE). EALE uses the Sobel operator for local feature extraction, and MTGE has a cross-attention mechanism to facilitate information interaction. To integrate information from both paths selectively and adaptively, we design a bilateral gated transformer module (BGTM) to combine deep semantic information effectively. Experiments on three segmentation tasks—skin lesions, polyps, and brain tumors—show that the proposed DPUSegDiff outperforms other state-of-the-art (SOTA) methods in segmentation performance and generalization ability. The code has been released on GitHub (https://github.com/Fanyyz/DPUSegDiff).

We aimed to address the issues of low encryption complexity and insufficient resistance to statistical attacks in chaotic image encryption algorithms. This paper proposes a novel method that combines hash scrambling, and bit-level transformation, leveraging the synergy between chaotic mapping and spatial-frequency transformation. Firstly, a specific hash function is constructed using the cubic chaotic system to achieve efficient scrambling of the image. Subsequently, the frequency domain data is generated by the Fourier transform and represented in the form of an optimized complementary code. Finally, the encryption is completed by DNA primary diffusion, secondary diffusion based on adjacent blocks, and the neighbor bit deprivation operation, and the entire process is jointly regulated by the Chen chaotic system and the Sin-Tent-Cos chaotic system. The experiment demonstrates that this scheme exhibits excellent robustness and security, with all indicators surpassing those of traditional methods. In particular, its resistance to differential attacks is close to the theoretical optimal value, demonstrating the important application potential of spatial-frequency joint encryption.