Analysis of Chronic Wound Images Using Factorization Based Segmentation and Machine Learning Methods

In this paper, an attempt has been made to perform an accurate assessment of chronic wound images. Pressure, venous and arterial leg ulcers are considered in this study. For this purpose, chronic wound images acquired by digital camera are enhanced using color correction, noise removal and color homogenization. Enhanced images in Cb color channel of YCbCr color space is used to extract wound bed with factorization based segmentation approach. Binary classification is performed to classify pressure ulcers and leg ulcers. The obtained results showed that the proposed segmentation method is capable of converging exactly to irregular wound boundaries. Hence, the suggested pipeline of processes seems to be promising for automatic segmentation and classification of pressure ulcers from leg ulcers aiding in the assessment of wound healing status.