Implementation of AIC based on I-frame only coding in H.264 and comparison with other still frame...

17
Implementation of AIC based on I- frame only coding in H.264 and comparison with other still frame image coding standards such as JPEG, JPEG 2000, JPEG-LS, JPEG-XR Radhika Veerla [email protected] EE Graduate student, UT Arlington MULTIMEDIA PROCESSING LAB ELECTRICAL ENGINEERING DEPARTMENT, THE UNIVERSITY OF TEXAS AT ARLINGTON Presentation for Ambrado Inc. Richardson, Texas, on June 13, 2008

Transcript of Implementation of AIC based on I-frame only coding in H.264 and comparison with other still frame...

Implementation of AIC based on I-frame only coding in H.264 and

comparison with other still frame image coding standards such as

JPEG, JPEG 2000, JPEG-LS, JPEG-XR

Radhika Veerla [email protected] Graduate student,

UT Arlington

MULTIMEDIA PROCESSING LAB ELECTRICAL ENGINEERING DEPARTMENT, THE UNIVERSITY OF TEXAS AT ARLINGTON

Presentation for Ambrado Inc. Richardson, Texas, on June 13, 2008

Advanced Image Coding Block Diagram

(a) Encoder [1] (b) Decoder [1]

Advanced Image Coding It is a still image compression system which is a combination of H.264

and JPEG standards.Features: No sub-sampling- higher quality / compression ratios 9 prediction modes as in H.264 Predicted blocks are predicted from previously decoded blocks Uses DCT to transform 8x8 residual block instead of transform

coefficients as in JPEG Employs uniform quantization Uses floating point algorithm Coefficients transmitted in scan-line order Makes use of CABAC similar to H.264 with several contexts

Proposed AIC Algorithm

(a) Proposed AIC Encoder

(b) Proposed AIC Decoder

BG

R

CrCbYCC

Mode Selectand Store

BlockPredict

modeY

Y, Cb, Cr Blks

+

+

Pred B

lk

FDCT Q ZZ Huff

AAC

Q1IDCT+

Table

Res

Res

Dec

Y Dec

YDec

Cb

Dec

Cr

Predictor

ModeEnc

BG

R

CrCbYICC

BlockPredict

Y,Cb,Cr Blks

++P

red Blk

IDCT Q1 IZZ IHuff

AADTable

Res

ModeDecand Store

mode

DecYDecCbDecCr

CC - color conversion, ICC - Inverse CC, ZZ – zig-zag scan, IZZ – inverse ZZ, AAC – adaptive arithmetic coder, AAD – AA decoder.

H.264 Block Diagram

[2]

H.264 Main Profile Intra-Frame Coding

•Transform block size reduced from 8x8 to 4x4•H.264 relies on spatial prediction taking the advantage of inter-block spatial correlation•Uses multiplier-less integer transforms and implemented in 16-bit fixed point architectures•Block DCT with inter-block correlation is competitive with global wavelet coding used in JPEG2000•Use CABAC or CAVLC

H.264 High Profile Intra-Frame Coding

•H.264 Fidelity range extensions support higher-resolution color spaces

•Advantage- improves coding efficiency by adding 8x8 integer transform, prediction schemes associated with adaptive selection between 4x4 and 8x8 transforms

JPEG Encoder and Decoder

(a) Encoder [6]

(b) Decoder [6]

JPEG-Baseline•8x8 block based DCT•Scalar quantization •Different quantization tables for luminance and chrominance components•Huffman coding

JPEG2000•Relies on wavelet transform•EBCOT scheme for coding wavelet coefficients•Adaptive context-based binary arithmetic coding•This project disables tiling and scalable mode for comparison as they adversely affect rate-distortion performance

Evaluation MethodologySoftwares and parameters used for

comparison

Standards Software Parameter Setting

AIC AIC reference quality

JPEG JPEG-Baseline Ref.

quality

H.264 JM software quantization

JPEG 2000 JasPer rate

JPEG-XR JPEG-XR ref. quality

Codec SettingsH.264• Main and high profiles in 4:2:0 coding mode• Activate intra coding profile for Frext• Activate RGB coding mode• 8x8 transform mode: enabled, allowing adaptive choice between (4x4) /(8x8)

transform and all associated prediction modes• Motion estimation: enabled• CABAC: enabled• R-D optimization: enabled • De-blocking filter: enabled

HD photo• No tiling• One-level of overlap in the transformation stage• No color space sub-sampling• Spatial bit-stream order• All sub-bands are included without any skipping

Image Quality Measures

• Criteria to evaluate compression quality• Two types of quality measures Objective quality measure- PSNR, MSE Structural quality measure- SSIM• MSE and PSNR for a NxM pixel image are defined as

where x is the original image and y is the reconstructed image. M and N are the width and height of an image and ‘L’ is the maximum pixel value in the NxM pixel image.

M

m

N

n

nmynmxNM

MSE1 1

2,,*

1 (1)

MSE

LPSNR

2

10log10 (2)

Structural Similarity Method• This method emphasizes that the Human Visual System (HVS) is

highly adapted to extract structural information from visual scenes. Therefore, structural similarity measurement should provide a good approximation to perceptual image quality.

• The SSIM index is defined as a product of luminance, contrast and structural comparison functions. [14]

where μ is the mean intensity, and σ is the standard deviation as a round estimate of the signal contrast. C1 and C2 are constants. M is the numbers of samples in the quality map.

Results for same resolutions

(a) Lena (512x512x24) (b) Airplane (512x512x24)

(c) Peppers (512x512x24) (d) Sailboat (512x512x24)

0 2 4 6 8 10 1215

20

25

30

35

40

45

50

55

60

65

Bits Per Pixel

PS

NR

dB

Quality vs Compression

AIC

AIC-HuffmanAIC-Adapt-AC

JPEG-Ref

JPEG

JPEG2000H.264-Main

H.264-High

0 1 2 3 4 5 6 7 8 9 1015

20

25

30

35

40

45

50

55

60

Bits Per Pixel

PS

NR

dB

Quality vs Compression

AICAIC-Huffman

AIC-Adapt-AC

JPEG-Ref

JPEG

JPEG2000H.264

0 2 4 6 8 10 12 1410

20

30

40

50

60

70

Bits Per Pixel

PS

NR

dB

Quality vs Compression

AIC

AIC-Huffman

AIC-Adapt-AC

JPEG-Ref

JPEG

JPEG2000H.264

0 5 10 1510

15

20

25

30

35

40

45

50

55

Bits Per Pixel

PS

NR

dB

Quality vs Compression

AICAIC-Huffman

AIC-Adapt-AC

JPEG-Ref

JPEG

JPEG2000H.264

Results (contd.)

(e) Splash (512x512x24) (f) Couple (256x256x24)

(g) Cameraman (256x256x8) (h) Man (256x256x8)

0 1 2 3 4 5 6 7 8 9 1010

20

30

40

50

60

70

Bits Per Pixel

PS

NR

dB

Quality vs Compression

AICAIC-Huffman

AIC-Adapt-AC

JPEG-Ref

JPEG

JPEG2000H.264

0 2 4 6 8 10 1215

20

25

30

35

40

45

50

55

60

Bits Per Pixel

PS

NR

dB

Quality vs Compression

AICAIC-Huffman

AIC-Adapt-AC

JPEG-Ref

JPEG

JPEG2000H.264-Main

0 1 2 3 4 5 610

20

30

40

50

60

70

80

90

Bits Per Pixel

PS

NR

dB

Quality vs Compression

AIC-Huffman

JPEG-RefJPEG2000

H.264

0 0.5 1 1.5 2 2.5 3 3.5 4 4.510

20

30

40

50

60

70

80

Bits Per Pixel

PS

NR

dB

Quality vs Compression

AIC-Huffman

JPEG-RefJPEG2000

H.264

Results for different resolutions (contd.)

(i) Lena (32x32x24) (j) Lena (64x64x24)

(l) Lena (256x256x24) (k) Lena (128x128x24)

0 5 10 15 20 2515

20

25

30

35

40

45

50

55

60

Bits Per Pixel

PS

NR

dB

Quality vs Compression

AIC-Huffman

JPEG-RefJPEG2000

H.264

0 2 4 6 8 10 12 14 16 1815

20

25

30

35

40

45

50

55

60

Bits Per Pixel

PS

NR

dB

Quality vs Compression

AIC-Huffman

JPEG-RefJPEG2000

H.264

0 2 4 6 8 10 12 1415

20

25

30

35

40

45

50

55

60

Bits Per Pixel

PS

NR

dB

Quality vs Compression

AIC-Huffman

JPEG-RefJPEG2000

H.264

0 2 4 6 8 10 12 1420

30

40

50

60

70

80

Bits Per Pixel

PS

NR

dB

Quality vs Compression

AIC-Huffman

JPEG-RefJPEG2000

H.264

Conclusions and Future Work• AIC outperforms JPEG by about 5dB and performs similar to or surpasses

the JPEG2000 performance below 2bpp.• Typical bit rates for AIC are 0-2bpp for color images and 0-4bpp for gray

scale images • H.264 outperforms every other codec for images of all resolutions, but

works close to other codecs in case of gray-scale images. The main concern is its complexity.

• For gray scale images, all the codecs including H.264 have similar performance.

• The gap between various standards increases with decrease in image resolutions.

• The limitation of JPEG reference software is that it has low dynamic range.• AIC is preferred because of its optimal performance with reduced

complexity and increased speed.• Comparison with JPEG-LS and JPEG-XR and also including SSIM

distortion measurement in Rate-Distortion curves (PSNR and SSIM vs bpp) can be the future work.

References[1] AIC website: http://www.bilsen.com/aic/[2] T. Wiegand, G. Sullivan, G. Bjontegaard and A. Luthra, “Overview of the H.264/AVC Video Coding Standard ”, IEEE

Transactions on Circuits and Systems for Video Technology, vol. 13, pp.560-576, July 2003[3] P. Topiwala, “Comparative study of JPEG2000 and H.264/AVC FRExt I-frame coding on high definition video

sequences,” Proc. SPIE Int’l Symposium, Digital Image Processing, San Diego, Aug. 2005.[4] P. Topiwala, T. Tran and W.Dai, “Performance comparison of JPEG2000 and H.264/AVC high profile intra-frame

coding on HD video sequences,” Proc. SPIE Int’l Symposium, Digital Image Processing, San Diego, Aug. 2006.

[5] T.Tran, L.Liu and P. Topiwala, “Performance comparison of leading image codecs: H.264/AVC intra, JPEG 2000, and Microsoft HD photo,” Proc. SPIE Int’l Symposium, Digital Image Processing, San Diego, Sept. 2007.

[6] G. K. Wallace, “The JPEG still picture compression standard,” Communication of the ACM, vol. 34, pp. 31-44, April 1991.

[7] H.264/AVC reference software (JM 12.2) Website: http://iphome.hhi.de/suehring/tml/download/[8] JPEG reference software Website: ftp://ftp.simtel.net/pub/simtelnet/msdos/graphics/jpegsr6.zip[9] JPEG2000 latest reference software (Jasper Version 1.900.0) Website: http://www.ece.ubc.ca/mdadams/jasper[10] Microsoft HD photo specification: http://www.microsoft.com/whdc/xps/wmphotoeula.mspx[11] D. Marpe, V. George, and T.Weigand, “Performance comparison of intra-only H.264/AVC HP and JPEG 2000 for

a set of monochrome ISO/IEC test images”, JVT-M014, pp.18-22, Oct. 2004.[12] M.J. Weinberger, G. Seroussi and G. Sapiro, “The LOCO-I lossless image compression algorithm: principles and

standardization into JPEG-LS”, IEEE Trans. Image Processing, vol. 9, pp. 1309-1324, Aug.2000.http://www.hpl.hp.com/loco/

[13] G. J. Sullivan, “ ISO/IEC 29199-2 (JpegDI part 2 JPEG XR image coding – Specification),” ISO/IEC JTC 1/SC 29/WG1 N 4492, Dec 2007

[14] Z. Wang and A. C. Bovik, “Image quality assessment: from error visibility to structural similarity,” IEEE Trans. Image Processing, vol. 13, pp. 600 – 612, Apr. 2004.