Photoacoustic Microscopy (PAM) imaging combines the high contrast of optical imaging with the deep penetration of ultrasound imaging, offering the advantages of high resolution and depth imaging. However, light scattering and ultrasound attenuation during the PAM imaging process limit imaging depth and resolution. In addition, environmental and system noise can affect the quality and stability of photoacoustic signals, resulting in artifacts and noise in the images. While previous research has primarily focused on photoacoustic image reconstruction, studies on PAM image enhancement remain relatively scarce. In this paper, we propose an algorithm based on the Hybrid Activation and Half Normalization (HAIN) block and a multi-stage U-Net. We also designed a Supervised Multi-Attention (SMA) module to connect the two stages. By combining channel attention and pixel attention, and incorporating ground truth supervision, the SMA module effectively extracts crucial global and detailed information. Experiments show that our proposed HAINet achieved an average peak signal-to-noise ratio of 37.38 dB and a structural similarity  of 0.972 in the PAM image enhancement task. HAINet also outperformed the comparative experiments in the PAM image denoising task.
Photoacoustic Microscopy (PAM) imaging combines the high contrast of optical imaging with the deep penetration of ultrasound imaging, offering the advantages of high resolution and depth imaging. However, light scattering and ultrasound attenuation during the PAM imaging process limit imaging depth and resolution. In addition, environmental and system noise can affect the quality and stability of photoacoustic signals, resulting in artifacts and noise in the images. While previous research has primarily focused on photoacoustic image reconstruction, studies on PAM image enhancement remain relatively scarce. In this paper, we propose an algorithm based on the Hybrid Activation and Half Normalization (HAIN) block and a multi-stage U-Net. We also designed a Supervised Multi-Attention (SMA) module to connect the two stages. By combining channel attention and pixel attention, and incorporating ground truth supervision, the SMA module effectively extracts crucial global and detailed information. Experiments show that our proposed HAINet achieved an average peak signal-to-noise ratio of 37.38 dB and a structural similarity  of 0.972 in the PAM image enhancement task. HAINet also outperformed the comparative experiments in the PAM image denoising task.
 

Photoacoustic microscopy imaging,multi-stage U-Net,hybrid activation,supervised multi-attention