Video Encoding

The ENCODE class of functions takes raw frames as input and compresses them into a bitstream.

Input frames usually come encoded in a repeated pattern called the Group of Picture (GOP) sequence. For example, a GOP sequence can start from an I-frame, followed by a few B-frames, a P-frame, and so on. ENCODE uses an MPEG-2 style GOP sequence structure that can specify the length of the sequence and the distance between two key frames: I- or P-frames. A GOP sequence ensures that the segments of a bitstream do not completely depend upon each other. It also enables decoding applications to reposition the bitstream.

ENCODE processes input frames in two ways:

Display order: ENCODE receives input frames in the display order. A few GOP structure parameters specify the GOP sequence during ENCODE initialization. Scene change results from the video processing stage of a pipeline can alter the GOP sequence.

Encoded order: ENCODE receives input frames in their encoding order. The application must specify the exact input frame type for encoding. ENCODE references GOP parameters to determine when to insert information such as an end-of-sequence into the bitstream.

An ENCODE output consists of one frame of a bitstream with the time stamp passed from the input frame. The time stamp is used for multiplexing subsequent video with other associated data such as audio. The SDK library provides only pure video stream encoding. The application must provide its own multiplexing.

ENCODE supports the following bitrate control algorithms: constant bitrate, variable bitrate (VBR), and constant Quantization Parameter (QP). In the constant bitrate mode, ENCODE performs stuffing when the size of the least-compressed frame is smaller than what is required to meet the Hypothetical Reference Decoder (HRD) buffer (or VBR) requirements. (Stuffing is a process that appends zeros to the end of encoded frames.)

 

For more complete information about compiler optimizations, see our Optimization Notice.