Positron Emission Tomography (PET) plays a vital role in oncological imaging by capturing the metabolic activity of tissues. However, ultra-low-dose PET (ULD-PET) scanning-designed to minimize radiation exposure-often produces images with substantial noise and diminished diagnostic reliability. To address this issue, a Conditional Wavelet Diffusion Model (cWDM) is applied for denoising ULD-PET images. This approach formulates denoising as a conditional image-to-image translation task, wherein clean PET images are reconstructed from their noisy counterparts. The cWDM integrates wavelet-domain features along with wavelet features extracted from ULD-PET images as conditional inputs, enabling the model to more effectively capture structural details and noise characteristics. The framework is trained and evaluated using paired full-dose and ULD-PET images. Experimental results demonstrate that the application of cWDM surpasses existing denoising methods in both noise reduction and structural fidelity, underscoring its potential for improving ULD-PET imaging in clinical applications.