Learned Image Compression with Wavelet Preprocessing for Low Bit Rates
Sofia Iliopoulou, Panagiotis Tsinganos, Dimitris Ampeliotis, Athanassios N. Skodras · 2023
Deep Learning has revolutionized the field of image processing and image compression in particular. A lot of research has been done in recent years on the subject of learning-based image compression, which has resulted in methods with increased compression performance but high computational complexity. The most successful methods eradicate the redundancies by using entropy modelling. In this paper, we utilize the Discrete Wavelet Transform (DWT) as a preprocessing step for a simple hyperprior model. The proposed method is compared to both traditional and deep learning-based techniques. It proves to be superior in lower bit rates, using both Peak Signal-to-Noise Ratio (PSNR) and Multi-Scale Structural Similarity (MS-SSIM) as metrics for the evaluation of the model.