Opened 11 years ago

Closed 11 years ago

#34 closed enhancement (fixed)

Configurability of scanning method and context model (as 0.7+)

Reported by: wjhan Owned by:
Priority: minor Milestone:
Component: HM Version:
Keywords: Cc: hurumi@…, fbossen, ksuehring, davidf, jct-vc@…

Description

TMuC supports a new context modeling for the transform coefficients and an adaptive scanning methods. But now they are controlled by a single macro, HHI_TRANSFORM_CODING, although they are different coding tools and can be separated easily.

We suggest to separate one macro into two macros: HHI_TRANSFORM_SCAN and HHI_TRANSFORM_CONTEXT_MODEL to improve the configurability of the software. Or, to have a backward compatibility, adding HHI_TRANSFORM_SCAN to turn on/off adaptive scanning inside the original MACRO is also ok.

Attachments (1)

patch_Ticket34.patch (2.2 KB) - added by wjhan 11 years ago.
On/off of adaptive scanning macro is added

Download all attachments as: .zip

Change History (11)

comment:1 follow-up: Changed 11 years ago by TN

The adaptive scan is linked with the context modeling stage for sig flag and context initialization of both sig and last flag. Therefore I do not recommend the separation.

comment:2 in reply to: ↑ 1 Changed 11 years ago by ilkoo.kim@…

Replying to TN:

The adaptive scan is linked with the context modeling stage for sig flag and context initialization of both sig and last flag. Therefore I do not recommend the separation.

If so, is there any way to check the performance of adaptive scan only without coupling with the new context modelig, or vice versa ?

comment:3 follow-up: Changed 11 years ago by TN

Of course you can use the adaptive scan with another context modeling as in TMuC with HHI_TRANSFORM_CODING = 0. But the results could be different as expected, because the current implementation is a combination of context modeling and adaptive scan (for sig and last flags).

comment:4 Changed 11 years ago by fbossen

  • Type changed from defect to enhancement

comment:5 in reply to: ↑ 3 Changed 11 years ago by wjhan

  • Cc hurumi@… added

Replying to TN:

Of course you can use the adaptive scan with another context modeling as in TMuC with HHI_TRANSFORM_CODING = 0. But the results could be different as expected, because the current implementation is a combination of context modeling and adaptive scan (for sig and last flags).

You may be right. Adaptive scan with another context model may bring some side-effects, but the other direction is somewhat clear to me. Context model with another scan. It is possible to use typical zigzag with the use of new context model, which means, HHI_TRANSFORM_CODING = 1, but adaptive scan is turned off. It adds more configurability to the software. Simply HHI_TRANSFORM_SCAN macro can be additionally defined when the HHI_TRANSFORM_CODING macro is defined. You can cross-verify our relevant patches for the confirmation.

comment:6 Changed 11 years ago by anonymous

The reason for the separation of the adaptive scan is not clear to me. Your argument is the improve configurability. From this point of view, there are a lot of parts that can made configurable too. I think TracTicket is not a right place to start a discussion. Please use email in case of reply (tung@…).

Changed 11 years ago by wjhan

On/off of adaptive scanning macro is added

comment:7 Changed 11 years ago by fbossen

  • Resolution set to fixed
  • Status changed from new to closed

Fixed in r249

comment:8 Changed 9 years ago by davidf

  • Component set to HM

Updating component after adding WD (Text) tickets

comment:9 Changed 9 years ago by davidf

  • Cc fbossen ksuehring davidf added

comment:10 Changed 9 years ago by davidf

  • Cc jct-vc@… added
Note: See TracTickets for help on using tickets.

This list contains all users that will be notified about changes made to this ticket.

These roles will be notified: Reporter, Owner, Subscriber, Participant

  • David Flynn(Subscriber, Participant)
  • Frank Bossen(Subscriber, Participant)
  • ilkoo.kim@…(Participant)
  • jct-vc@…(Subscriber)
  • karl.sharman@…(Always)
  • Karsten Suehring(Subscriber, Always)
  • Woo-Jin Han(Reporter, Subscriber, Participant)