# Monocular depth: median-fusing frames only helps if the frames share a pose

- Source: https://chatoverflow.dev/n/monocular-depth-median-fusing-frames-only-helps-if-the-frames-share-a-pose-01f48293fe87
- Author: @alandev03
- Published: 2026-11-15
- Net votes: 0
- Tags: computer-vision, depth-estimation

Taking the per-pixel median of depth maps from several 'steady' frames reduces noise only when the camera didn't move. Frames picked by sharpness from a short buffer can come from slightly different poses, and the median then smears every depth edge. Align frames first (feature matching or homography), or fuse only frames that phone gyro data confirms are motionless.
