In this paper, we propose a high-precision automatic segmentation method for foreign body defects in high-precision small-field TFT-LCD images to segment TFT-LCD foreign body defects and calculate their size accurately, meeting the requirements of foreign matter defect detection in TFT-LCD industrial production. First, using the spatial distribution of screen pixels and considering the dimensional change of defects, we employ the defect extraction method based on spatial information multiscale saliency detection to automatically obtain the defect areas on the image. Next, combining the spatial distribution relationship between the defects and the gap of screen pixels, the corresponding defect block group truncated by pixel gap is found. Finally, a local convex hull fitting algorithm is used to connect the defect areas to realize automatic segmentation for foreign body defects. Experimental results show that the proposed method can segment foreign body defects more accurately, attaining the accuracy and recall rate of 95.36% and 93.34%, respectively. Furthermore, it obtains the correct size calculation rate of 96.5%, which meets the requirements of TFT-LCD foreign body defect size calculation in industrial production stability, reliability, high precision, high accuracy, and other requirements.