SRI developed a novel binarization technique for extracting text from gray-scale images of scanned mail pieces or other documents. Conventional binarization methods may perform poorly if the background is not uniform, if the contrast between the foreground and background is small or variable, or if noise is present in the image. SRI's technique, specifically designed to work with images of text, can accommodate local variations in intensity without picking up extraneous interference. In addition, it can detect printed characters with poor contrast (Figure 1).