As an observer of video codec development, I heard Daala (PVQ) used activity masking to make quantization steps finer (don't discard all detail) in low-contrast areas. I don't know if that's practical in AV1 (a DCT block-based codec).