• State Resolved
  • Locked Locked
  • Replies 3 replies
  • Subscribers 98 subscribers
  • Views 936 views
  • Users 0 members are here
Support feedback
Options
Options
Related

This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

Range of motion vector/SAD values?

JNoring
Prodigy 230 points


What is the possible range of motion vector and SAD values from TI's H.264 encoder (i.e. this guy: http://tiexpressdsp.com/index.php/Using_MV_SAD_Info_of_DM365_encoder_in_Application)?  And are the vectors longer in one direction than another?  In other words, can the horizontal components have a greater range that the vertical components?

This question was previously asked on the forums (see http://e2e.ti.com/support/dsp/davinci_digital_media_processors/f/100/p/33881/119987.aspx ), but I don't understand the answer.

 

Thanks in advance,

Jeremy

  • 0
    TI__Genius 17215 points

    The range of motion vector in horizontal dircetion can be +- 29.75 and in vertical direction can be even more than this as H.264 encoder supports vertical GMV(the max limit will be based on the level limit as specified in the standard).

    What is the real problem which you are facing ? 

    regards

    Yashwant

  • 0 in reply to Yashwant Dutt
    Prodigy 230 points

    Yashwant,

     

    Thanks for the reply.  I'm working on using this information for motion detection, which is why the range is important (particularly if I'm using the motion vectors; if there's a difference in range, that will skew MD in one direction or another).  Is the quoted range in the horizontal in qpel units, or pixel units?  Because I definitely see larger values than that come out of the encoder.  We're encoding with high profile, level 3.1, if that helps you at all.

    Lastly, I notice that my vectors in the horizontal direction (i.e. MVx) are typically much larger than in the vertical direction (i.e. MVy), so I'm not sure what to make of your response.  I typically see my vectors in the MVy direction max out around ~50 (in qpel units, so the actual info directly from the encoder), while my MVx component can be 100+.  If I visualize the vectors, it's very clear that in the vertical, vectors are shorter than in the horizontal.

    Any comments on this?

     

  • 0 in reply to JNoring
    TI__Genius 17215 points

    The range which i provided is in full pel unit, the one returned by MV SAD is in quarter pel. It may be possble to have more horizontal MV compared to vertical MV. It may be due to the nature of the motion. Also, the vertical search range itself is less compared to horizontal. But there is support of vertical GMV which may increase it, but may not happen always.

    regards

    Yashwant