The tool wear condition in micro-milling significantly affects the geometry and surface quality of critical parts in miniature components, which are key factors for product quality and performance stability. However, due to the small size of the tools, real-time monitoring of tool wear is challenging in actual machining processes and severely impacts machining efficiency. This paper proposes a tool wear prediction method based on fractal features of machined surface images. The method primarily utilizes multifractal analysis to extract various texture features from machined surface images and constructs a dataset with the actual tool wear values obtained from an image acquisition system. Then, the tool wear is predicted using a tool wear evaluation technique that combines feature selection algorithms and support vector regression (ReliefF–SVR). The results show that the proposed method exhibits strong robustness under various cutting conditions and can accurately predict tool flank wear in the micro-milling process, with an average prediction accuracy of over 93.6%. This study presents a feasible and reliable solution for the actual quality control of micro-precision parts.