Components made of polymeric composite materials, such as wings and stabilizers of aircrafts, are periodically inspected using non-destructive testing methods. Ultrasonic testing is one of the primary inspection methods applied in the aircraft/aerospace industry. Image processing methods have been developed for the purpose of ultrasonic data analysis and increasing the inspection efficiency. A critically important factor in damage sizing is appropriate processing of ultrasonic data in order to extract the damage region properly before calculation of its extent. The paper presents a comparative analysis of various image segmentation methods in the light of accuracy of damage detection in ultrasonic C-Scans of composite structures. A brief review of image segmentation methods is presented and their usefulness in the ultrasonic testing applications is discussed. The selected methods, namely the threshold-, edge-, region-, and clustering-based ones, were tested using ultrasonic C-Scans of a specimen with barely visible impact damage and aircraft panels with delamination, all made of CFRP composites. A problem with the selection of appropriate input parameters using most of the segmentation methods is discussed, and non-parametric histogram-based approaches are proposed. A quantitative analysis of the accuracy of damage detection using the segmentation methods is presented and the most suitable approaches are introduced. The proposed processing procedures may significantly improve the objectivity of inspections of composite elements and structures using ultrasonic testing.