Detection of Cloth from Human Images with Marker Controlled Watershed Segmentation and Morphological Analysis

Fashion image segmentation is a growing field of research. It plays an important role in e-commerce applications such as apparel recognition, search and retrieval, and automatic recommendation. The major challenges in clothing image classification include the large variety of styles, textures, and colors; pose variations; uncontrolled lighting; diverse designs; image blurring; and skin ambiguities. This paper discusses the use of Marker-Controlled Watershed Segmentation to detect textile regions in input images. The approach is divided into two phases: first, marking the foreground using morphological operations, and second, marking the background using thresholding operations. Finally, the watershed segmentation algorithm is applied to locate the cloth region. Experiments were conducted on a synthetic dataset of 50 human images using MATLAB for evaluation.