ICIP 2008 Review Results for paper #2075

혹시나 했더니 역시나~, reject!!

Thank you for submitting the above listed manuscript #2075 for consideration in the ICIP 2008 program. We regret to inform you that your paper has not been accepted for presentation at the conference. The reviewers' comments on your paper are attached at the end of this email. These comments served as part of the basis for the Technical Program Committee's decision and may offer useful feedback that can be used to strengthen future publication submissions.

The Technical Program Committee carefully selected reviewers for ICIP 2008 and assigned reviewers papers in their areas of expertise. This year we received over 1700 papers and were only able to accept about 45%.

While we recognize that this information is disappointing to you, we would encourage you to attend the conference and meet with authors whose papers are in your field of interest. Full information regarding ICIP 2008 will be posted on the conference Web site, http://www.icip08.org, and will list the complete technical program shortly.

We look forward to welcoming you in San Diego.

Sincerely,

Antonio Ortega and B.S. Manjunath
ICIP 2008 Technical Program Co-Chairs


---- Comments from the Reviewers: ----
The paper proposes to use SVM to quickly decide between Intra and Inter coding in H.264/AVC. The results show a significant speed-up, for a negligible coding efficiency loss.

English needs improvement.

"To choose the best predicted mode for a macroblock, the H.264/AVC utilizes Lagrangian rate-distortion optimization", it should state "The H.264/AVC JM" or "JM11.0" instead of "H.264/AVC".

"PSNR-Y degradation from 0 to 0.2 dB" maybe "0 to 0.02 dB"

-----
In this paper, the authors proposed an inter/intra mode selection algorithm for H.264/AVC video coding as a pattern classification approach using Support Vector Machine (SVM). The SVM classifier is incorporated into the mode inter/intra mode decision process with the feature vectors of SATD values for 16*16 Intra, 4*4Intra and 16*16 Inter modes.

Here are some suggestions for further improve the quality of the paper:

1. There are some spelling mistakes.
2. Some labels are missed in Table 1.
3. The experimental results in Table 2 are lack of specific experimental conditions, i.e. the value of QP
4. Table 3 and Table 4 are more suitable to put in section 3.2 to explain why the authors choose the linear model for kernel.

-----
The paper investigates inter/intra mode selection in H.264/AVC. In stead of using the computationally expensive rate-distortion optimization based mode selection method adopted in the reference codec of H.264, the authors propose to use a support vector machine-based classier to achieve fast mode selection. However, with the exception of using support vector machine, a learning theoretic (e.g. classification) approach to mode selection in H.264 has been studied quite extensively in

A. Jagmohan and K. Ratakonda, "Time-efficient learning theoretic algorithms for H.264 mode selection," in Proc. ICIP 2004.

It seems that the authors are not aware of this important reference, where in addition to inter/intra mode selection, selection among prediction modes for intra blocks, were also studied. Moreover, experimental results seem to suggest the method proposed in this paper performs no better than that in the above reference for inter/intra mode selection.

-----