Subsampling Text Images

For digital display of scanned documents containing text there is often the need to reduce the image size so that it can be completely viewed. To do this, filtering and subsampling are required. We present three methods for image size reduction where the objective in each is to retain high readability. The three subsampling methods are: binary to gray-scale, binary to binary, and a non-uniform, adaptive binary to binary method where non-text regions are reduced preferentially to preserve text regions. These methods are tested, the parameter values giving the best results are given, and the limits in terms of image reduction versus readability are discussed.