Blur is the loss of fine detail and texture - grass turning to a green smear, skin losing pores, brick losing mortar. It is the quiet member of the block-transform family, caused when the quantizer discards the high-frequency DCT information that carries fine detail, or when the video was downscaled and never recovered. It is also the flip side of deblocking: when a modern encoder runs out of bits, its in-loop deblocking filter smooths the unavoidable error into softness instead of seams, which is why heavily compressed modern streams look mushy where older video would have looked tiled - blocking is traded for blur. The hard part for both eye and metric is that blur is not always an artifact: a shallow depth of field is intentional softness, so the skill is telling a creative choice from a compression failure. SSIM and VMAF detect blur reasonably well because it is a structural, spatial change, but neither can tell intentional softness from a fault on its own.
The handbook · 7 volumes
Fora Soft Video Engineering Handbook
Blur is hard to catch because some softness is intentional, so teams need metrics that separate lost detail from creative choice. We wrote seven volumes covering the whole video stack, and for blur start with Volume 1: quantization and detail loss, scaling, VMAF and other perceptual metrics in H.264, HEVC and AV1, with Volume 6 on AI enhancement. Written by Nikolay Sapunov, CEO at Fora Soft.
Viewers calling your video soft while your metrics say it is fine?
Share a few encodes with the sources: we measure detail loss and blur per scene and tell you which encoder or scaling settings to change. 250+ video products shipped since 2005.
