Study of various motion estimation algorithms in H.264 &

implementation of an optimal algorithm in LabVIEW

Presented by:SRIRAM B

11MCE0006

Internal Guide: ESWAR REDDY M(Asst. Professor, VIT)External Guide: Shri. Anand.K & Smt.Subha Varier (VSSC,ISRO)

School of Electronics EngineeringVIT University, Vellore-632014

Objective of Project

• To study various motion estimation techniques for H.264 based compression and selection of an optimal algorithm.

• Coding of the various techniques in MATLAB for comparitive analysis.

• Implementation of the selected technique in LabVIEW for hardware implementation.

H.264 video CODEC Introduction• H.264/MPEG-4 Part 10 or AVC (Advanced Video Coding) is a standard for video

compression, and is currently one of the most commonly used formats for the

recording, compression, and distribution of high definition video. • officially released in Sept, 2005. • developed by the partnership effort of ITU-T Video Coding Experts Group

(VCEG) together with the ISO Moving Picture Experts Group (MPEG)[Joint Video

Team (JVT)]

H.264 Encoder

Motion Estimation Algorithms1. Full Search Algorithm (Exhaustive Search Algorithm)

Mean Absolute Difference:

2. Three Step Search Algorithm

3. Advanced Three Step Search Algorithm

4. Simple and Efficient Three Step Search Algorithm

If MAD(A) ≥ MAD(B) & MAD(A) ≥ MAD(C), select (b) If MAD(A) ≥ MAD(B) & MAD(A) ≤ MAD(C), select (c)If MAD(A) < MAD(B) & MAD(A) < MAD(C), select (d) If MAD(A) < MAD(B) & MAD(A) ≥ MAD(C), select (e)

5. Four Step Search Algorithm

6. Diamond Search Algorithm

LDSP SDSP

7. Adaptive Rood Pattern Search Algorithm

The predicted motion vector for previous block is (3,-2) , and the step size S = Max( |3|, |-2|) = 3.

• Homogenous movement of objects utilized here.

Evaluation criterian1. PSNR: PSNR in dB is given by

PSNR=10log10(I2/MSE)

Where I is the maximum intensity level MSE – Mean Square Error

A - Original Image B - Reconstructed Image M & N - size of image

2. Structural Similarity Metric Index (SSIM)

A=Original Image B=Reconstructed Image μA & μB – mean intensities of data A & B

σA & σ B – standard deviations of data A & B

3. Average of no. of search points per macroblock for the algorithm.

Simulated results

Caltrain

football

mrchest

Advantages of ARPS algorithm over other algorithms

• No. of computations least among the algorithms preserving the PSNR & SSIM values.

• If the predicted motion vector is (0, 0), it does not waste computational time in doing LDSP, it rather directly starts using SDSP.

• Furthermore, if the predicted motion vector is far away from the center, then again ARPS save on computations by directly jumping to that vicinity and using SDSP.

Due to the reasons mentioned above, Adaptive Rood Pattern Search Algorithm is selected as the optimal algorithm for any type of application.

LabVIEW output

Conclusions

• Studied 7 different various motion estimation techniques for H.264 based compression and Comparative study was carried out for all seven algorithms in MATLAB for assessment to the three analytical parameters namely PSNR, SSIM and Number of search points. Based on this comparative study, Adaptive Rood Pattern Search algorithm is chosen as the optimal algorithm for different applications taken and was taken up for implementation• For programming on the target FPGA, LabVIEW was chosen for implementation. Adaptive Rood Pattern Search Algorithm was implemented and the performance indices were compared.• The results obtained in MATLAB and LabVIEW shows a close match.

• Gate level implementation has to be done for the selected algorithm ‘Adaptive Rood Pattern Search Algorithm’ in LabVIEW.

• Hardware implementation can be done in LabVIEW’s processor cum FPGA device called CompactRIO.

• Also with the help of this selected algorithm, a healthy H.264 standard can be integrated which can be utilized for a lot of H.264 compression/coding based applications.

Future scope

References[1] Iain E.Richardson ,“The H.264 Advanced video compression standard” (Second edition), 2010, John Wiley &

Sons, Ltd.[2] Study materials and tutorials from website. http://www.vcodex.com[3] T.Wiegand, G.J.Sullivan, G.Bjontegaard, A.Luthra, “ Overview of the H.264/AVC video coding standards”, IEEE

transactions on circuits and systems for video technology, Vol.13, No.7, July 2003.[4] Faizul Hadi Jamil, Ali Chekima, Rosalyn R. Porle, Othman Ahmad, Norfarariyanti Parimon, “BMA

Performance of Video Coding for Motion Estimation”, IEEE 2012 Third International Conference on Intelligent Systems Modelling and Simulation, 978-0-7695-4668-1/12 DOI 0.1109/ISMS.2012.115.

[5] Renxiang Li, Bing Zeng, and Ming L. Liou, “A New Three-Step Search Algorithm for Block Motion Estimation”, IEEE Trans. Circuits And Systems For Video Technology, vol 4., no. 4, pp. 438-442, August 1994.

[6] SUN Ning-ning, FAN Chao, XIA Xu, “An Effective Three-step Search Algorithm for Motion Estimation”, 2009 IEEE, 978-1-4244-3929-4/09.

[7] Jianhua Lu, and Ming L. Liou, “A Simple and Efficent Search Algorithm for Block-Matching Motion Estimation”, IEEE Trans.Circuits And Systems For Video Technology, vol 7, no. 2, pp. 429-433,April 1997.

[8] Lai-Man Po, and Wing-Chung Ma, “A Novel Four-Step Search Algorithm for Fast Block Motion Estimation”, IEEE Trans. Circuits And Systems For Video Technology, vol 6, no. 3, pp. 313-317, June 1996.

[9] Shan Zhu, and Kai-Kuang Ma, “ A New Diamond Search Algorithm for Fast Block-Matching Motion Estimation”, IEEE Trans. Image Processing, vol 9, no. 2, pp. 287-290, February 2000.

[10] Yao Nie, and Kai-Kuang Ma, “Adaptive Rood Pattern Search for Fast Block-Matching Motion Estimation”, IEEE Trans. Image Processing, vol 11, no. 12, pp. 1442-1448, December 2002.

[11] Study materials and tutorials from website. http://www.ni.com.[12] Nasser Kehtarnavaz, Sidharth ahotra,” Digital Signal Processing Laboratory: LabVIEW - based FPGA

Implementation”, Universal-Publishers, 2010

Thank you