x265 1.8版本更新

0.  2015.10.8 更新1.8版本(git时间不对居然提早了16天，骗子 )
     支持12位   avx2指令集 优化   熵编码优化   添加一些新工具

    

x265 version 1.8 has been released. This release supports 12bit input

depths, a large amount of AVX2 optimizations, entropy coding optimizations,

as well as new quality features.

 1.8 improves 1.23% sizewise, IOW x265 1.8 encodes slightly more efficiently than v1.7. I haven't compared the resulting video clips yet.

Edit: v1.8 output is quite different however x265 still blurs everything out vs. x264 at comparable bitrate (which is in my case circa 3.2Mbit/sec for 720p@30fps quite static very dark video containing a lot of noise). 

1.  删除1/2分辨率的mvc SAD cost计算

   r emove mvc's sad cost calc for lowres,     already measured in slicetype   

2. 跳过ME，在MVC的cost为0的情况

   bypass the ME search if any of MVC Cost   is Zero

3.  rcbug

    fixes inconsistent output in linux because of RC   Lock issue in CQP/CRF

4. dia修正

    improve Diamon Search by reduce duplicate   SAD

5. mvcost检错

    fix overflow in mvcost() check failure

6. 场景切换帧修改  

    Modify Scenecut algorithm to detect scene  transition points

7. 添加 lamda表格

    added lambda tables based based on qp values

8. rc修正

   fix rate factor calculation after updating  m_avgQpRc in rateControlEnd (Divya Manivannan)

9. ffmpeg scale 缩放问题     

   Hi Zsolt.

 

    This is not an x265 issue, specifically.

 

   You may have been lucky that your original dimensions returned an even

    number as height when you scaled to a width of 1280 pixels. But when you

    scaled to a width of 1920 pixels, your height would possibly have to be

    scaled to an odd number, which is not allowed in YUV 4:2:0 subsampling.

 

   I don't know how to force swscaler in ffmpeg to return at least "MOD 2"

   results; but x265 is not to be blamed.

10. cutree多次计算duration问题

    

On 08/14, Ximing Cheng wrote:

> Hi, the source code in lookahead cuTree function has to calculate the

> averageDuration by a while loop each time, why not use double

> averageDuration = (double)m_param->fpsDenom / m_param->fpsNum; in the

> initial function of lookahead? Thanks!

 

The cu-tree code was adapted from x264's mb-tree and x264 supports

variable frame rates. x265 doesn't support VFR yet and so you're right

we could save some cycles there.

 

At some point in the future we might add VFR support to x265 so you have

to weigh possible future work (adding it back) against the compute

savings.

11. 支持4:0:0格式   

  Today's Topics:  

  Re: [PATCH] add support for Monochrome color space

      (X265_CSP_I400) (Steve Borho)

真正的qstep是pow（2，(qp-4)/6.0）,量化x就是用(x+offset)/qstep

12. 支持Android  20150929

To subscribe or unsubscribe via the World Wide Web, visit

https://mailman.videolan.org/listinfo/x265-devel

or, via email, send a message with subject or body 'help' to

x265-devel-request@videolan.org

 

You can reach the person managing the list at

x265-devel-owner@videolan.org

 

When replying, please edit your Subject line so it is more specific

than "Re: Contents of x265-devel digest..."

 

 

Today's Topics:

 

   1. Re: [PATCH] cost: fix mode cost check (Steve Borho)

   2. [PATCH] Experimental ARMv8-A Crosscompilation (Peter Kov??)

   3. [PATCH] Experimental Android support (Peter Kov??)

   4. [PATCH] Experimental ARMv7-A Crosscompilation (Peter Kov??)

13. log信息打印  finishFrameStats函数中剔除了每帧信息打印  转到csv中

14. scenecut算法更新

15. 添加265_AQ_AUTO_VARIANCE_BIASED

x2651.8缺少的工具：

1） 广义B帧
2)   时域MV
3)  LDC(slice->m_bCheckLDC 无发现其用途)
4）只对list0和list1的第一帧判断加权与否，其它的都设定为不加权
5）RateControlEntry类中bframes 无任何作用
7)  RateControl::rateEstimateQscale  rce->sliceType = m_sliceType; 冗余代码
8)  rce->leadingNoBSatd = m_leadingNoBSatd; 冗余代码  leadingNoBSatd 貌似无任何意义
9)  collectCTUStatistics函数中 统计NxN模式存在问题 ctu.m_partSize[absPartIdx] < AMP_ID 改为 ctu.m_partSize[absPartIdx] <= AMP_ID 合适
10) FrameEncoder 中m_endFrameTime 无任何意义
11) getPUBelowLeftAdi 获取左下参考像素  同一CU下不能获取左下参考块（存在风险）
12)checkDefaultScalingList() 函数 存在风险  应该是相同返回true