TY - GEN
T1 - GABIC
T2 - 31st IEEE International Conference on Image Processing, ICIP 2024
AU - Spadaro, Gabriele
AU - Presta, Alberto
AU - Tartaglione, Enzo
AU - Giraldo, Jhony H.
AU - Grangetto, Marco
AU - Fiandrotti, Attilio
N1 - Publisher Copyright:
© 2024 IEEE
PY - 2024/1/1
Y1 - 2024/1/1
N2 - While standardized codecs like JPEG and HEVC-intra represent the industry standard in image compression, neural Learned Image Compression (LIC) codecs represent a promising alternative. In detail, integrating attention mechanisms from Vision Transformers into LIC models has shown improved compression efficiency. However, extra efficiency often comes at the cost of aggregating redundant features. This work proposes a Graph-based Attention Block for Image Compression (GABIC), a method to reduce feature redundancy based on a k-Nearest Neighbors enhanced attention mechanism. Our experiments show that GABIC outperforms comparable methods, particularly at high bit rates, enhancing compression performance.
AB - While standardized codecs like JPEG and HEVC-intra represent the industry standard in image compression, neural Learned Image Compression (LIC) codecs represent a promising alternative. In detail, integrating attention mechanisms from Vision Transformers into LIC models has shown improved compression efficiency. However, extra efficiency often comes at the cost of aggregating redundant features. This work proposes a Graph-based Attention Block for Image Compression (GABIC), a method to reduce feature redundancy based on a k-Nearest Neighbors enhanced attention mechanism. Our experiments show that GABIC outperforms comparable methods, particularly at high bit rates, enhancing compression performance.
KW - Attention Mechanism
KW - Graph Neural Network
KW - Learned Image Compression
KW - Vision Transformer
UR - https://www.scopus.com/pages/publications/85211646544
U2 - 10.1109/ICIP51287.2024.10647413
DO - 10.1109/ICIP51287.2024.10647413
M3 - Conference contribution
AN - SCOPUS:85211646544
T3 - Proceedings - International Conference on Image Processing, ICIP
SP - 1802
EP - 1808
BT - 2024 IEEE International Conference on Image Processing, ICIP 2024 - Proceedings
PB - IEEE Computer Society
Y2 - 27 October 2024 through 30 October 2024
ER -