부산대학교 도서관

The significance of enhancing image compression efficiency for machine vision, analysis, and comprehension tasks has gained increasing recognition. In response to this need, we propose and implement a novel method called ELIC (Efficient Learned Image Compression with Unevenly Grouped Space-Channel Contextual Adaptive Coding) to achieve high compression efficiency. Our method is evaluated on the classic OpenImage V6 Common Test Condition (CTC) eval datasets, and its performance is compared to baseline methods for machine vision tasks. The results of our study demonstrate a substantial enhancement in compression efficiency, suggesting that the ELIC technique holds promise for pushing the boundaries of state-of-the-art visual compression for vision tasks. Furthermore, we believe that our approach can promote the application of learning-based image compression.