Variable-rate learned image compression with integer-arithmetic-only inference

doi:10.1016/j.jvcir.2025.104634

Variable-rate learned image compression with integer-arithmetic-only inference

2025-11-04

0

PRE

AI

DOI：10.1016/j.jvcir.2025.104634

Original

Original request for help

Save

Abstract

En 中文

• We propose the first fully quantized variable-rate learned image compression model. • We design a calibration procedure for variable rates with different parameters. • We adopt reduced accumulator precision for efficient hardware implementation.

Keywords:

Integer-arithmetic inference

Learned image compression

Model quantization

Variable-rate

Variable-rate learned image compression with integer-arithmetic-only inference

Abstract

Journal

Organization