Abstract: Infrared thermography effectively detects internal defects in metals, non-metals, and composite materials. Defect size is a key parameter for evaluating the impact of defects. We used a half width height measurement algorithm to achieve semi-automatic measurement of defect size, manually drawing a straight line through the center of the defect to form a spatial pixel curve, and used the SG filtering algorithm to filter, automatically finding the half width height position of the spatial pixel curve, thus achieving defect size measurement. Through research on ABS plastic, carbon steel, and carbon fiber composite material specimens, it was found that the defect sizes measured by infrared images at different times have different errors. Using clear infrared images at different times, the measurement error was within 10%, whereas using fuzzy infrared images at different times, the measurement error was approximately 20%. This will effectively improve the accuracy of defect size measurements.