A generalized interpolative vector quantization method for jointly optimalquantization, interpolation, and binarization of text images

This paper presents an approach for the effective combination of interpolat ion with binarization of gray level text images to reconstruct a high resol ution binary image from a lower resolution gray level one. We study two non linear interpolative techniques for text image interpolation, These nonline ar interpolation methods map quantized low dimensional 2 x 2 image blocks t o higher dimensional 4 x 4 (possibly binary) blocks using a table lookup op eration. The first method performs interpolation of text images using conte xt-based, nonlinear, interpolative, vector quantization (NLIVQ), This syste m has a simple training procedure and has performance (for gray-level high resolution images) that is comparable to our more sophisticated generalized interpolative VQ (GIVQ) approach, which is the second method. In it, we jo intly optimize the quantizer and interpolator to find matched codebooks for the low and high resolution images. Then, to obtain the binary codebook th at incorporates binarization with interpolation, we introduce a binary cons trained optimization method using GIVQ, In order to incorporate the nearest neighbor constraint on the quantizer while minimizing the distortion in th e interpolated image, a deterministic-annealing-based optimization techniqu e is applied. With a few interpolation examples, are demonstrate the superi or performance of this method over the NLIVQ method (especially for binary outputs) and other standard techniques e.g., bilinear interpolation and pix el replication.