16
An efficient Video Coding using Phase- matched Error from Phase Correlation Information Manoranjan Paul1 and Golam Sorwar2 2008 IEEE

An efficient Video Coding using Phase-matched Error from Phase Correlation Information Manoranjan Paul 1 and Golam Sorwar 2 2008 IEEE

Embed Size (px)

Citation preview

Page 1: An efficient Video Coding using Phase-matched Error from Phase Correlation Information Manoranjan Paul 1 and Golam Sorwar 2 2008 IEEE

An efficient Video Coding using Phase-matchedError from Phase Correlation Information

Manoranjan Paul1 and Golam Sorwar2

2008 IEEE

Page 2: An efficient Video Coding using Phase-matched Error from Phase Correlation Information Manoranjan Paul 1 and Golam Sorwar 2 2008 IEEE

Outline• Introduction• Propose technique– Phase correlation– Binary Matrix Generation using Phase Correlation– Mode Selection from the Binary Matrix

• Computational complexity• Simulation results• Conclusion

Page 3: An efficient Video Coding using Phase-matched Error from Phase Correlation Information Manoranjan Paul 1 and Golam Sorwar 2 2008 IEEE

Introduction• Variable block size motion estimation and

compensation in H.264. How to choose mode?(RDO)

• Some Fast mode selection mode sacrifice some quality or increase bit streams to reduce computation time.

• Here we advise a algo. using phase correlation technique to predict block size saving 50% time without degrading the image quality.

Page 4: An efficient Video Coding using Phase-matched Error from Phase Correlation Information Manoranjan Paul 1 and Golam Sorwar 2 2008 IEEE

Phase correlation• In image processing, phase correlation is a method of image

registration, and uses a fast frequency-domain approach to estimate the relative translative offset between two similar images.

Example:

Page 5: An efficient Video Coding using Phase-matched Error from Phase Correlation Information Manoranjan Paul 1 and Golam Sorwar 2 2008 IEEE

• To obtain the Phase Correlation of two images, perform these steps:

1. DFT:2. cross power spectrum:3. IDFT:4. Determine the location of the peak in

Phase correlation

Page 6: An efficient Video Coding using Phase-matched Error from Phase Correlation Information Manoranjan Paul 1 and Golam Sorwar 2 2008 IEEE

two motions

one motion no motion

(5,6)

(5,7)(6,5)

Page 7: An efficient Video Coding using Phase-matched Error from Phase Correlation Information Manoranjan Paul 1 and Golam Sorwar 2 2008 IEEE

Binary Matrix Generation using Phase Correlation

• Phase-matched error 1. )2. )3. )4. )5. ) r:reference bolck, c:current blocke:phase-mached error

Page 8: An efficient Video Coding using Phase-matched Error from Phase Correlation Information Manoranjan Paul 1 and Golam Sorwar 2 2008 IEEE

Binary Matrix Generation using Phase Correlation

[13] To L., M. Pickering, M. Frater, and J. Arnold, “A motion confidence measure from phase information,” IEEE International Conference on Image Processing, pp. 2583-2586, 2004.

Page 9: An efficient Video Coding using Phase-matched Error from Phase Correlation Information Manoranjan Paul 1 and Golam Sorwar 2 2008 IEEE

Binary Matrix Generation using Phase Correlation

• The ratio D represents the proportion of energy in the lower-frequency components.

6. )

7. )

Page 10: An efficient Video Coding using Phase-matched Error from Phase Correlation Information Manoranjan Paul 1 and Golam Sorwar 2 2008 IEEE

Binary Matrix Generation using Phase Correlation

T=0.3 green blocks (8x8 pixels) indicate motion and other areas indicate no motion.

Page 11: An efficient Video Coding using Phase-matched Error from Phase Correlation Information Manoranjan Paul 1 and Golam Sorwar 2 2008 IEEE

Mode Selection from the Binary Matrix

• The resolution of an image is HxW pixels.• The size of the binary matrix will be H/8xW/8.• 16x16-pixel block is considered as a video

encoding processing unit, thus we will get four sub-blocks.

Based on the binary matrix, we will decide which mode it would be for ME&MC.

Page 12: An efficient Video Coding using Phase-matched Error from Phase Correlation Information Manoranjan Paul 1 and Golam Sorwar 2 2008 IEEE

Mode Selection from the Binary Matrix

Page 13: An efficient Video Coding using Phase-matched Error from Phase Correlation Information Manoranjan Paul 1 and Golam Sorwar 2 2008 IEEE

(a)&(b)proposed technique (c)&(d)H.264 cyan,green, blue, and red color for 16x16, 16x8, 8x16, and 8x8 and others respectively.

Page 14: An efficient Video Coding using Phase-matched Error from Phase Correlation Information Manoranjan Paul 1 and Golam Sorwar 2 2008 IEEE

Computational Complexity• Full search motion estimation using d width

search length.– H.264 kH(3N2(2d+1)2) , kH :average modes per MB – Proposed technique kP(3N2(2d+1)2+5.25N2)

• (kH- kP)/kH : result half number modes per MB at middle range of bit rates.

Page 15: An efficient Video Coding using Phase-matched Error from Phase Correlation Information Manoranjan Paul 1 and Golam Sorwar 2 2008 IEEE

Simulation results

Page 16: An efficient Video Coding using Phase-matched Error from Phase Correlation Information Manoranjan Paul 1 and Golam Sorwar 2 2008 IEEE

Conclusion & Future Work• We proposed a video coding technique using phase

correlation information to select motion estimation and compensation modes, and reduces around 50% computational time without losing any image quality compared to the H.264.

• In furture search, using phase correlation to generate

motion vector and seeing the effect of threshold on different bit rates is necessary.