We propose an algorithm for separating the foreground (mainly text and linegraphics) from the smoothly varying background in screen content images. Theproposed method is designed based on the assumption that the background part ofthe image is smoothly varying and can be represented by a linear combination ofa few smoothly varying basis functions, while the foreground text and graphicscreate sharp discontinuity and cannot be modeled by this smooth representation.The algorithm separates the background and foreground using a least absolutedeviation method to fit the smooth model to the image pixels. This algorithmhas been tested on several images from HEVC standard test sequences for screencontent coding, and is shown to have superior performance over other popularmethods, such as k-means clustering based segmentation in DjVu and shapeprimitive extraction and coding (SPEC) algorithm. Such background/foregroundsegmentation are important pre-processing steps for text extraction andseparate coding of background and foreground for compression of screen contentimages.
展开▼