Mammogram Compression Using Super-Resolution

As mammography moves towards completely digital and produces prohibitive amounts of data, compression plays an increasingly important role. Although current lossless compression methods provide very high-quality images, their compression ratios are very low. On the other hand, several lossy compression methods provide very high compression ratios but come with considerable loss of quality. In this work, we describe a novel compression method that consists of downsampling the mammograms before applying the encoding procedure, and applying super-resolution techniques after the decoding procedure to recover the original resolution image. In our experiments, we examine the tradeoffs between compression ratio and image quality using this scheme, and show it provides significant improvements over conventional methods.