[On working][Help needed]The codecs page

encoding/video-encoding.md

@@ -1,10 +1,149 @@
 # Video Encoding
 
-<TODO - mention codecs, new and old>
-<TODO - this whole page>
-
-This page is under construction.
-To help finish this page,
-you can contribute by clicking
-<kbd><i class="fa fa-edit"></i> EDIT THIS PAGE</kbd>
-at the top left.
+It's a long history while we talking about video codecs.
+
+![Evolution of video coding standards](images/standard_history.png)
+
+Nowadays, while h.264 is no doubt
+the most popular video coding format
+all over the world,
+you can still find mpeg2
+in some DVD video sources.
+With the help of these video codecs,
+we can easily compress
+a big raw video into a tiny one.
+
+What's more,
+we should distinguish video codec from
+the video coding format.
+Although video coding formats
+such as H.264/HEVC are sometimes
+referred to as codecs,
+there is a clear conceptual difference
+between a specification
+and its implementations.
+Video coding formats are
+defined in specifications,
+while video codecs are
+the software that implements
+these specifications.
+For example,
+the video codec X264 is
+the implementation of
+the video coding format H.264.
+
+In the next chapter,
+we'll talk about some methods
+during the compression
+and the parameters that
+control these processes.
+
+## X264
+
+**
+Please trying to not explain this in complex way.
+**
+<TODO - Motion estimation(--me,--subme,--me-range)>
+<TODO - Intra prediction>
+<TODO - Definition of I,P,B Frame(--keyint,--min-keyint,--bframes,--b-adapt,--scenecut)>
+<TODO - Multiple Reference Frame(--ref)>
+<TODO - DCT Transform>
+<TODO - Quantization coding(I do not know what it exactly called)(--crf,--qp,2pass)>
+<TODO - Video Buffering Verifier(VBV)(--vbv-bufsize,--vbv-maxrate)>
+<TODO - Macroblock Tree(--no-mbtree,--qcomp,--rc-lookahead,--qpmin,--qpmax)>
+<TODO - Adaptive Quantizers(--aq-mode,--aq-strength,--trellis,--chroma-qp-offset)>
+<TODO - Deblock(--deblock)>
+<TODO - (not in H264, it is a encoder-side optimizations)Psychological optimization(I do not know what it exactly called)(--psy-rd)>
+
+### Motion Estimation
+
+Almost every video we process today
+has much redundant information.
+This is to say,
+for many frames of a video,
+the only difference between one frame
+and another is the result of either
+the camera moving or an object
+in the frame moving.
+This indicates that much of the
+information renders one frame will be
+the same as the used in the following frame.
+So we don't need to keep all the
+information in each frame.
+Instead, we can store a frame
+carries full information, 
+and the subsequent frames can
+only save the different part. 
+
+
+![A group of pictures](images/motionestimation_1.png)
+
+![Difference between those pictures detected by the codec](images/motionestimation_2.png)
+
+H264 defines three ways 
+
+Here's some commonly used options in X264 that affect the motion estimation stage while encoding:
+
+`--me <str>` decides which algorithm X264 will use in motion estimation
+- diamond: `--me dia` is the simplest search,
+consisting of starting at the best predictor,
+checking the motion vectors at one pixel
+upwards, left, down, and to the right,
+picking the best, and repeating the process
+until it no longer finds any better motion vector.
+It's a very fast but bad choice,
+I'll only recommend this if
+your computer is less powerful.
+
+- hex: `--me hex` consists of a similar strategy,
+except it uses a range-2 search
+of 6 surrounding points,
+thus the name.
+It's a bit slower than diamond
+but still not an ideal choice,
+both hex and dia should only
+be used if you are live-streaming.
+
+- multiple hexagon: `--me umh` **Recommended!**
+UMH searches a complex multi-hexagon pattern
+in order to avoid missing
+harder-to-find motion vectors.
+This can bring some pretty decent result
+while still having pretty fast speed,
+choose this if you have
+limited computing power
+or the time is important for you.
+
+- exhaustive: `--me esa` is mathematically equivalent
+to the bruteforce method of searching
+every single motion vector in the area. 
+It's slower than "umh" and
+the result isnot as good as "tesa".
+I'll consider this as a
+moderate choice but not recommended.
+
+- SATD exhaustive: `--me tesa` **Recommended!**
+It's an algorithm which attempts to
+approximate the effect of running
+a Hadamard transform comparison
+at each motion vector;
+like exhaustive, but a little bit better
+and a little bit slower.
+If you want the best result,
+this is for you.
+
+`--subme <int 0-10>` set the subpixel estimation complexity.
+Higher numbers are better.
+The recommend value of this parameter is 10.
+If you want to get a bit faster, 8 is a good choice.
+
+`--me-range <int default:16>` decides the max range
+of the motion search in pixels. 
+Higher numbers are better but slower.
+It's recommended to set 24 for 720p content
+and 36 for 1080p content.
+
+`--no-fast-pskip` Disables early skip detection on P-frames.
+Very slightly increases quality at some speed cost.
+It depends on you to use it or not.
+