Image harmonization plays an important role in computer vision, enhancing the realism of composite images. However, existing work focus on color adjustments while neglecting the impact of texture on color coherence. To address this issue, we propose the Texture and Color Dynamic Network (TCDNet), a new dual-encoder single-decoder architecture. Our TCDNet aims to achieve image harmonization through a unified texture-color perspective from both foreground and background regions. Specifically, we employ two task-specific encoders, i.e., a texture encoder and a color encoder, to separately extract texture and color features. Subsequently, we designed a Texture based Color Transfer (TBCT) module to align the color representation of the foreground with that of the background, leveraging texture-based cues. Within TBCT, attention mechanisms and position encoding refine textural details, ensuring consistent texture alignment of foreground and background. During decoding, we propose a Color Dynamic (CoDy) module to dynamically adapt kernels to navigate and reinforce color correlations across varying input conditions. This synergistic interplay between texture and color dynamics enables TCDNet to navigate the complex landscape of image harmonization with high precision. We conducted extensive experiments on synthetic and real data to demonstrate the competitive performance of our method when compared to state-of-the-art (SOTA) supervised approaches.