Traditional Transformation Theory Guided Model for Learned Image Compression
This work addresses ultra low bitrate image compression, a domain-specific problem, with incremental improvements by integrating traditional transformations into deep learning methods.
The paper tackles the problem of ultra low bitrate image compression, which is underexplored compared to medium and high bitrates, by proposing an invertible encoding network guided by traditional transformation theory, resulting in a codec that outperforms existing methods in compression and reconstruction performance.
Recently, many deep image compression methods have been proposed and achieved remarkable performance. However, these methods are dedicated to optimizing the compression performance and speed at medium and high bitrates, while research on ultra low bitrates is limited. In this work, we propose a ultra low bitrates enhanced invertible encoding network guided by traditional transformation theory, experiments show that our codec outperforms existing methods in both compression and reconstruction performance. Specifically, we introduce the Block Discrete Cosine Transformation to model the sparsity of features and employ traditional Haar transformation to improve the reconstruction performance of the model without increasing the bitstream cost.