The rapid evolution of deep neural networks in computer vision has driven unprecedented performance gains, yet their growing complexity poses critical deployment barriers in latency-sensitive scenarios like autonomous systems and edge devices. Knowledge distillation (KD) emerges as a pivotal paradigm to bridge this gap by transferring knowledge from cumbersome teacher models to compact student models. This chapter introduces the background of KD, providing a comprehensive overview of its motivation, evolution, methodologies, and applications in computer vision. It sets the foundation for understanding the systematic approaches and innovations discussed throughout this book.