Divx Compression – The Encoder
DVD Videos use MPEG-2 standard of encoding data to cater to the need for higher quality video although it comes at a price of larger disk space requirement. The MPEG-4 standard picked up because of its ability and flexibility to compress data and maintain video quality at the same time. This has been exceedingly helpful for media that needs to be streamed online. One can even take a backup of 5~6 movies on One single DVD. This standard is being used by different Codecs (Coder-Decoder) like Divx, XviD and 3ivx to treat digital video in a way so as to maintain quality while increasing packing strength. Divx compression technique follows the MPEG-4 standard. This standard basically specifies how digital video should be encoded. The Codec pack follows these basic principals and also uses its own algorithms, special techniques for better quality and compression.
Let us see how Divx operates:
1) Frames and Macroblocks
A video consists of frames. Each frame is further divided in to 16 x 16 pixel squares known as “Macroblocks”. Each Macroblock in a frame is compared to the previous (or forward) frame. If motion is detected in the frame, then it is represented by a vector (has direction as well as magnitude). This “motion vector” represents the position of the image in a Macroblock (with respect to the previous frame). Referring to previous frames for motion estimation allows the codec to store less information of the entire Macroblock. There might be a slight difference between the original and the encoded form. Divx makes up for this with its own data correction during Playback.
2) I, P and B-Frames
Each frame is treated in a different way depending on the amount of motion detected in a frame. There are three ways MPEG-4 may encode a frame.
If very little or no change is detected from the previous frame then it is encoded as an Intra-frame (I-Frame). This means that the Macroblocks store the information as an image, rather than a Motion Vector. I-frames are also known as “Key-frames”. Increasing the number of I-frames increases the quality and the size of the final Output.
Predicted Frames (P-Frames) use motion estimation as discussed above. They usually follow a “key-frame” and there can be several in a row. The blocks in a P-frame include ones that have been predicted with the help of the vectors in the previous frame. Blocks in a P-Frame include ones that have been predicted with the help of vectors in a previous frame. Blocks that don't have any motion vectors assigned to them are saved as Images, similar to an I-frame.
Bi-directional (B-frames) may be present in an encoded video stream if the option is enabled during the encode process. These frames refer to the previous frame as well as the next one. A B-frame can also contain “Intra-blocks” – as mentioned above – if it can not forward or backward predict it. Bi-directional encoding produces a better compression ratio keeping quality in mind and is useful while creating low bit rate videos.
3) Spatial and Temporal Filtering
Spatial and Temporal Filtering are techniques to remove or normalize aberrations in color. Spatial Filtering involves the comparison of color between adjacent pixels in a single frame while temporal compares adjacent frames. These smoothen sharp changes in color between adjacent pixels or frames.
All the information stored for each pixel is further converted into frequency values (in an ascending order) using complex mathematical function. This conversion is used to increase the compression ratio with minimum loss in quality.
4) Quantizers
The “Quantizer” is fixed value that is used to control the accuracy of image data (according to the MPEG-4 standard). A higher value increases the percentage of error but also decreases the file size. For high quality “Quantization” one uses a lower image value.
TYPES OF ENCODING
1-Pass Encoding
A Single-pass encode means the video stream will be processed and compressed as it receives each frame. This is the fastest way to encode a Video Clip.
Multi-Pass Encoding
A Multi-Pass Encode can be a 1 st -pass or Nth pass. Both do more or less the same thing; the only difference being the number of times it analyzes the video clip. In both methods, the encoder goes through the entire video stream before beginning any sort of encoding. It decides on the bitrate as it goes through the Video and logs it in a file. If it is using 1 st -pass, it will begin encoding after the first analysis and use the log file for information on the frames. The nth-pass technique analyzes the clip a few times, updating the log file each time, fine-tuning it before the final encode. This procedure takes time but it produces an Optimal Quality Output.
Divx Playback – The Decoder
Software like Divx Player, QuickTime, VLC Media Player use the Divx Decoder to playback the encoded Video. The decoder is configured to correlate with the encoded Video stream and play the clip. It also has mechanism to smooth art-effects such as “deblocking” and “deringing”.
AVI format
The AVI format was originally created by Microsoft. It is a popular but dated format and at times creates heavier files. The “AVI” extension, however, has multi-dimension containing capability that supports varied codes like Divx, Cinepak, motion JPEG, XviD, etc. |
