Folks!
I want to know what will happen if I decide not to use official reference models e.g. JM, HM to implement new algorithms for possible publications?
I spent too much time in my Master's on working with JM for the sake of my thesis. But recently I found out that everything would work better if used x264, instead.
The problem is that I am not sure that if reviewers are OK with this choice? I can hardly remember a paper using x264/x265 as its reference model?
What do you think?
Hello,
You have to be aware that x264 is a commercial version of the AVC standard. Things run better on it since it does not have all the resources that can be found in H.264/AVC.
The H.264/AVC standard has all the video coding resources, that is why it is used by the "academic world". I guess you could use x264, however you should also compare to works that use this same standard version, and that is difficult.
Cheers!
Hi,
It is also important to remerber that, the JM implements all the specifications of the H264/AVC standard. However, if you expect that the HM does the same, actually it doesn't. For example, the intra module, only eight modes are calculated for a block, following an heuristic. So, if you need results using the full tools and specifications that the HEVC standard proposes, you will have to modify the reference software as well.
In addition, for cientific results, and better comparisons with other works, you should use the reference sofware, since it is used for almost all researches is this area.
I suppose it really depends on what you are intending to evaluate.
If your work is looking at any aspect of the encoding process you have to stick with the reference software to allow accurate comparisons. However if you are solely working on transmission issues you may be able to use x264/265.
We tend to always use the standard for any research work, but to use x264/265 where possible when building demos etc
From a practical point of view both encoders will produce an annexe B compliant bitstream. Therefore if only transmission is being considered it should not be an issue.
The only issue I can forsee is that of PSNR comparisons (or whatever other measure you use). If you use a comparison of the transmission between before and after your contribution I don't see an issue with that.
Mohsen, both encoders (HEVC and x265) generate a compliant bitstrem. However, you have to notice that x265 aims at market. Then, it doesn't implement all the features present in the HM reference software in order to be a lightweight solution. The coding efficiency results will also vary, since not the same tools will be used in the coding process.
Performing a quick research at IEEEXplore base, I've found only one paper which uses x265, and around 100 papers which use the HEVC Test Model (performing an abstract term research).
But, if you are comparing your results to other x26X papers, there is no problem of using it. But you if will compare the results with papers that use the reference sotware, it will sound strange and kind of unfair.
- Badges
- Science topic
- Similar topics
- Correction