Surface defect inspection is crucial for ensuring the quality stability of liquid crystal display (LCD) screens, particularly for TFT-LCDs.Known for its high efficiency and low cost, machine vision technology has become the primary means for inspecting TFT-LCD quality. This paper reviews the development of LCD and lists the common types of Mura defects. The traditional image processing methods and deep learning for detecting Mura defects are introduced, and recent advancements in image preprocessing techniques such as image filtering and brightness correction are summarized. The application of artificial intelligence techniques, such as supervised learning, unsupervised learning and transfer learning is introduced, in the detection of Mura defects on TFT-LCD surfaces. Finally, the research directions for machine vision-based Mura defect inspection technology on TFT-LCD are anticipated.