Half-frame in interlaced video — 50i or 60i splits each full frame into two temporally offset fields. Critical for deinterlacing and motion artifacts.
Field
When working with interlaced footage — 50i PAL or 60i NTSC — you split each full frame into two fields, which are temporally offset by half a frame period. The first field captures the odd lines, the second the even lines. Historically, this was the only method to transmit smooth motion with limited bandwidth. Today, where almost everything is progressive, you encounter interlaced material primarily in old broadcast archives, live TV recordings, or when clients still shoot with 50i cameras.
In the VFX pipeline, this becomes critical during deinterlacing. You fundamentally have three options: combine both fields (which causes motion blur), discard one field and upscale (loss of quality), or use intelligent interpolation that calculates new pixels between the fields. Some NLE systems offer real-time deinterlace filters — "Motion Adaptive" usually works best because it preserves static areas and interpolates moving areas. The wrong deinterlace mode can ruin your compositing work — shaky motion, combing artifacts on horizontal lines, lack of sharpness.
Practical tip: When working with old DV or HDV material, immediately check in the source format if it's truly interlaced. Some archives provide interlaced MOV files even though they contain progressive content — purely a storage convention. Before building effects, deinterlace in the sequence, not retroactively in the effect stack. Then, always continue working in 25p or 29.97p. If you have to capture yourself (rare, but occurs during archive digitization), ensure the field order is correctly recognized — "Upper Field First" or "Lower Field First" — otherwise, it will judder.
Another point: Some high-quality deinterlacing tools like YADIF or motion estimation integrated into Nuke are only necessary if you have truly action-heavy material. For talking-head archives, simple "Bob Deinterlacing" (both fields upscaled) is sufficient. The trade-off between processing time and visual quality quickly becomes relevant when you have to render thousands of frames.