Adapting Learned Image Codecs to Screen Content via Adjustable Transformations
This work addresses the bottleneck of learned image codecs for screen content compression, offering a backwards-compatible solution that is incremental in nature.
The paper tackled the problem of low coding efficiency of learned image codecs for screen content images by introducing parameterized linear transformations with neural prefilters and postfilters, achieving up to 10% bitrate savings compared to baseline methods while adding only 1% extra parameters.
As learned image codecs (LICs) become more prevalent, their low coding efficiency for out-of-distribution data becomes a bottleneck for some applications. To improve the performance of LICs for screen content (SC) images without breaking backwards compatibility, we propose to introduce parameterized and invertible linear transformations into the coding pipeline without changing the underlying baseline codec's operation flow. We design two neural networks to act as prefilters and postfilters in our setup to increase the coding efficiency and help with the recovery from coding artifacts. Our end-to-end trained solution achieves up to 10% bitrate savings on SC compression compared to the baseline LICs while introducing only 1% extra parameters.