Subsamping-based image compression prediction for memory-limited applications

ITU T.81 JPEG baseline compression is perhaps the most widely used image compression standard today. This standard, which is used extensively in Internet and image processing applications by a variety of OS platforms, has also found broad use in camera-equipped cellular phones. A limitation of the baseline standard is that, for a given set of quantization and Huffman encoding tables, the compressed size of an arbitrary image cannot be guaranteed. This can lead to problems in wireless applications where typical constraints include CPU capability, memory size and available data transfer bandwidth. Presented in this paper is a solution for overcoming these limitations 99% of the time for atypical-use cases, which takes advantage of frame-to frame similarity in a stream from a camera, and full-scale vs. subsampled image correlation for a given image.