0

I'm working a document, after read rfc6184 and Annex B confused with i p b frame with how many i frames or b p frames needed to decode a complete yuv or rgb frame(a picture can be im_write with opencv or something like n-dim array in numpy concept).

so the question is decode from h264 packet how many i frames, p frames, b frames do we need to make up a complete picture at least.

after read rfc6184 and Annex B, I can understand what i p b frames is, and nalu concept.

ald2000
  • 3
  • 2
  • What is your meaning of _"H264 packet"_ exactly? If we know, then we can tell you how many you need... PS: **(1)** Best to work at NALU level. Where each NAL (from this `00 00 01` **start code** to the next one) is either a full frame (check for type IDR) or is a slice that requires other slices (which can even from other frames) to construct a full image. **(2)** If you have P or B frames then you'll also **need the previous IDR keyframe** to construct a full image. Note: Keyframes can be two types: **IDR** (is full image like JPEG) or **I frame** (can be made from blocks of multiple frames) – VC.One Nov 13 '22 at 08:17
  • @VC.One thanks, i mean (2), when I received I B P frames, if keyframe is I frame(IDR can decode into JPEG) whether I must wait for how many B or P frames to construct a JPEG like I have got a IDR or even I do not need B or P frames, all I need is I keyframe. – ald2000 Nov 17 '22 at 11:43

0 Answers0