Remote Sensing (RS) image Change Detection (CD) serves as a crucial technical instrument across multiple domains including land use monitoring, urban expansion analysis, and resource exploration. However, current CD networks predominantly focus on bi-temporal feature interaction at identical scale and merely investigate multi-scale information fusion within individual network branch. To overcome the aforementioned challenges, we present an innovative spatio-temporal interaction network (STINet) with two key innovations. Firstly, we develop a spatio-temporal interaction module (STIM) that effectively aligns and enhances spatio-temporal consistency through bi-temporal feature interaction while integrating multi-scale semantic and detailed information. Subsequently, we design an adaptive pixel difference enhancement module (APDEM), which sequentially stacks a series of pixel difference complementary convolution groups (PDCGs) with different receptive fields to progressively capture subtle pixel-level changes from different perspectives, thus establishing a bidirectional and complementary feature extraction mechanism. Rigorous evaluations across three widely-recognized change detection datasets (SYSU-CD, WHU Building Dataset, and LEVIR-CD) validate STINet's exceptional efficacy. Relative to the sub-optimal network, STINet achieves IoU improvements of 1.72%, 1.19% and 0.38% on respective datasets, and its performance surpasses that of the current state-of-the-art (SOTA) methods. The implementation is publicly accessible at: https://github.com/ahaha-16/STINet .