Gel electrophoresis has significantly influenced the progress achieved in genetic studies over the last decade. Image processing techniques that are commonly used to analyze gel electrophoresis images require mainly three steps: band detection, band matching, and quantification and comparison. Although several techniques have been proposed to fully automate all steps, errors in band detection and, hence, in quantification are still important issues to address. In order to detect bands, many techniques were used, including image segmentation. In this paper, we present a novel, fully-automated technique based on the derivative of the standard deviation to perform segmentation and to detect protein bands. Results show that even for poor quality images with faint bands, segmentation and detection are highly accurate. Index Terms-Gel electrophoresis image, protein, band detection, segmentation.