Interim Report for EE 5359: Multimedia Processing

Interim Report for EE 5359: Multimedia Processing

Investigation of Image Quality of Dirac, H.264 and H.265

Biju Shrestha (UTA ID: 1000113697 Email: )

The University of Texas at Arlington

416 Yates Street, Arlington, Texas 76019-0016

Acronyms and Abbreviations

AVC / advanced video coding
BBC / British Broadcasting Corporation
CBR / constant bit rate
CODEC / coder and decoder
CSNR / channel signal to noise ratio
dB / decibel
FRExt / fidelity range extensions
FSIM / featured similarity index
GM / gradient magnitude
HEVC / high efficiency video coding
HVS / human visual system
IEC / international electrotechnical commission
ISO / international organization for standardization
IST / integer sine transform
ITU-T / international telecommunication union - telecommunication standardization sector
JPEG / joint photographic experts group
kbps / kilobits per second
LIVE / laboratory for image and video engineering
MICT / media information and communication technology laboratory
MPEG / moving picture experts group
MSE / mean squared error
MS SSIM / multi scale structural similarity metric
MSU / Moscow State University
PC / phase congruency
PSNR / peak signal to noise ratio
RGB / red, green and blue
SSIM / structural similarity metric
TID2008 / Tampere image database 2008
VBR / variable bit rate
VCEG / video coding experts group

Abstract

There exist several standards for video compression with additional improvements in performance and qualities in comparison to their older versions [2]. This project proposes to investigate the image quality of Dirac, H.264 and H.265 using metrics like PSNR, CSNR, MSE, SSIM, MS SSIM, and FSIM [3, 5, and 7] using various test sequences. The conventional metrics like PSNR and MSE are a measure of intensity and cannot measure the subjective fidelity [3]. This project report shows the progress made so far.

Introduction

Video codec is a tool which is used to compress and decompress the digital video [2]. There are several types of video compression methods. Few of them that are going to be discussed in this project are Dirac, H.264 and H.265 [1-3].

Dirac

Dirac video codec was initially developed by BBC Research [1]. It is an open source software project and is powerful and flexible despite using only small number of core tools [1]. The several features that Dirac offers are [1]:

·  Multi-resolution transforms

·  Inter and intra frame coding

·  Frame and field coding

·  Dual syntax

·  CBR and VBR operations

·  Variable bit depths.

·  Multiple chroma sampling formats

·  Lossless and lossy coding

·  Choice of wavelet filters

·  Simple stream navigation

Dirac has three main strands [15]. First is a compression specification for the byte stream and the decoder [15]. Second is software for compression and decompression and third are the algorithms designed to support simple and efficient hardware implementations [15]. Dirac despite being similar to many video coding systems had additionally adopted the combined effectiveness, efficiency and simplicity. The decoder and encoder architectures of Dirac are shown respectively in figures 1 and 2.

Figure 1. Dirac decoder architecture [18]

Figure 2. Dirac encoder architecture [15]

H.264

H.264 is also referred as AVC and it is a standard for video compression [2]. H.264/MPEG-4 AVC is one of the international video coding standards jointly developed by the VCEG of the ITU-T and the MPEG of ISO/IEC [11]. It provides enhanced coding efficiency for a wide range of applications like video telephony, video conferencing, TV, storage, streaming video, digital video authoring, digital cinema, etc. [11]. In addition, the FRExt provides enhanced capabilities relative to the base specification [11].

H.264 does not have a predefined CODEC but has the predefined syntax for decoding and encoding bit stream as shown in figures 3 and 4 respectively [1]. The various profiles of H.264 are shown in figure 5.

Figure 3. H.264 decoder [2]

Figure 4. H.264 encoder [2]

Figure 5. Various profile of H.264 [12]

H.265

H.265 is also known as HEVC [3] and it can deliver significantly improved compression performance relative to that of the AVC (ITU-T H.264 | ISO/IEC 14496-10) [10]. Alshina et al [16] investigated the coding efficiency with high resolution, HD 1080p, and concluded that it can be progressed by average 37% and 36% bit savings for hierarchical B structure and IPPP structure when compared to MPEG-4 AVC [16]. The typical block-based video codec is composed of many processes including intra prediction and inter prediction, transforms, quantization, entropy coding, and filtering [17] as shown in Figure 6. Over the decade, video coding techniques have gone through intensive research to achieve higher coding efficiencies [17].

Figure 6. Encoder block diagram of H.265. Grey boxes are proposed tools and white boxes are

H.264/AVC tools [17]

Figure 7. Decoder block diagram of H.265. Grey boxes are proposed tools and white boxes are H.264/AVC tools [27]

Image Quality Assessment using SSIM and FSIM

Digital images and videos are prone to different kinds of distortions during different phases like acquisition, processing, compression, storage, transmission, and reproduction [5]. This degradation results in poor visual quality. There are several metrics which are widely used to quantify the image quality like FSIM, SSIM, bitrates, PSNR and MSE [3, 8, 13, 14]. This project will primarily focus on metrics like SSIM, FSIM and bitrates. The other conventional metrics like PSNR and MSE will not be measured as they are directly dependent on the intensity of an image and do not correlate with the subjective fidelity ratings [3]. MSE cannot model the human visual system very accurately [4].The measured parameters like FSIM and SSIM of Dirac, H.264, and H.265 will be compared to study their comparative characteristics and make conclusions.

SSIM is the quality assessment of an image based on the degradation of structural information [5]. The SSIM takes an approach that the human visual system is adapted to extract structural information from images [14]. Thus, it is important to retain the structural signal for image fidelity measurement. Figure 8 shows the difference between nonstructural and structural distortions. The nonstructural distortions are changes in parameter like luminance, contrast, gamma distortion, and spatial shift and are usually caused by environmental and instrumental conditions occurred during image acquisition and display [14]. On the other hand, structural distortion embraces additive noise, blur, and lossy compression [14]. The structural distortions change the structure of an image [14]. Figure 9 explains the measurement system used in the calculation of SSIM.

Figure 8. Difference between nonstructural and structural distortions [14]

Figure 9. Block diagram of SSIM measurement system [5]

SSIM is based on the evaluation of three different metrics like luminance, contrast, and structure which are described mathematically by equations (1), (2), and (3) respectively [7].

/ ------(1)
/ ------(2)
/ ------(3)

Here,

µx and µy = local sample means of x and y respectively

σx and σy = local sample standard deviations of x and y respectively

σxy = local sample correlation coefficient between x and y

C1, C2, and C3 = constants that stabilize the computations when denominators become small

General form of SSIM index can be obtained by combining equations (1), (2) and (3) [7].

/ ------(4)

Here, α, β, and γ are parameters that mediate the relative importance of those three components. Using α = β = γ = 1. We get [7],

/ ------(5)

Figure 10 shows the different distorted images which are quantified using MSE and SSIM. It is clearly visible that the different images are of different quality based on human visual system (HVS). However, all the distorted images have approximately same MSE, whereas SSIM is less for poor quality image giving much better image quality indication than that of MSE.

(a)  Original
MSE = 0; SSIM = 1 /
(b)  Mean luminance shift
MSE = 144, SSIM = 0.988 /
(c)  Contrast stretch
MSE = 144, SSIM = 0.913

(d)  Impulse noise contamination
MSE = 144, SSIM = 0.840 /
(e)  Blurring
MSE = 144, SSIM = 0.694 /
(f)  JPEG compression
MSE = 142, SSIM = 0.662

Figure 10. MSE and SSIM measurement of images under different distortions. (a) original image, (b) mean luminance shift, (c) contrast stretch, (d) impulse noise contamination, (e) blurring, and (f) JPEG [22] compression [13]

FSIM is based on the fact that HVS understands an image mainly according to its low-level features [3]. PC is a dimensionless measure of the significance of a local structure [3]. PC and image GM measurements are used as primary and secondary feature respectively in FSIM [3]. FSIM score is calculated by applying PC as a weighting function on the image local quality characterized by PC and GM [3]. FSIM is designed for gray-scale images [3] and FSIMc incorporates the chrominance information. FSIM can be mathematically modeled as shown in equation 6 [3].

/ ------(6)

Here, SL(x) = overall similarity between reference image and distorted image

FSIMc can be mathematically modeled as shown in equation 7 and the computation process is illustrated in figure 11 [3].

/ ------(7)

Here, λ > 0 is the parameter used to adjust the importance of the chrominance components.

Figure 11. Illustration for FSIM/FSIMc index computation. f1 is the reference image, and f2 is a distorted version of f1 [3].

All the metrics use different approaches to compare the images quantitavely. This different approach makes one method different from another. Table 1 shows the ranking of image quality assessment metric performance on six databases. It can be seen from Table 1 that FSIM is better than SSIM and SSIM is better than PSNR when implementing an image quality assessment.

Table 1. Ranking of image quality assessment metrics performance (FSIM, SSIM and PSNR) on six databases [3].

TID2008 / CSIQ / LIVE / IVC / MICT / A57
FSIM / 1 / 1 / 1 / 1 / 1 / 1
SSIM / 2 / 2 / 2 / 2 / 2 / 2
PSNR / 3 / 3 / 3 / 3 / 3 / 3

Results

Video Information
QCIF sequence: foreman_qcif.yuv
Frame height: 176
Frame width: 144
Frame rate: 30 frame/second /
Figure 12: “foreman_qcif.yuv” [28]

Figure 13: PSNR achieved at various bitrates for foreman QCIF sequence using H.264 encoder

Figure 14: CSNR achieved at various bitrates for foreman QCIF sequence using H.264 encoder

Figure 15: MSE achieved at various bitrates for foreman QCIF sequence using H.264 encoder

Figure 16: SSIM achieved at various bitrates for foreman QCIF sequence using H.264 encoder

Figure 17: MS SSIM achieved at various bitrates for foreman QCIF sequence using H.264 encoder

Figure 18: PSNR achieved at various bitrates for foreman QCIF sequence using Dirac encoder

Figure 19: MSE achieved at various bitrates for foreman QCIF sequence using Dirac encoder

Figure 20: SSIM achieved at various bitrates for foreman QCIF sequence using Dirac encoder

Figure 21: MS SSIM achieved at various bitrates for foreman QCIF sequence using Dirac encoder

Conclusions

The project is aimed in studying the qualitative performances of different video codecs with a primary focus on Dirac, H.264 and H.265 [19 – 21]. Different parameters like PSNR, CSNR, MSE, SSIM, MS SSIM, and FSIM at various bitrates will be measured for all three video codecs to make a comparative study. Based on various test sequences of different spatial/temporal resolutions, MATLAB, Microsoft visual studio, and MSU video quality measurement tools [26] will be extensively used to perform image quality assessment of different codecs at various bit rates. Figures 13 to 17 shows the variation of metrics like PSNR, CSNR, MSE, SSIM, and MS SSIM respectively for various bitrates for foreman QCIF sequence using H.264 encoder. Further analysis is needed to be done using Dirac and H.265 encoder.

References

[1] Dirac Video (2008, September 23), “Dirac Specification” [Online]. Available: http://diracvideo.org/download/specification/dirac-spec-latest.pdf

[2] I. Richardson (2011), “A Technical Introduction to H.264/AVC” [Online]. Available: http://www.vcodex.com/files/H.264_technical_introduction.pdf

[3] L. Zhang, L. Zhang, X. Mou, and D. Zhang, “FSIM: A feature similarity index for image quality assessment,” IEEE Transactions on Image Processing, vol.20, no.8, pp.2378-2386, Aug. 2011.

[4] Z.Li and A.M. Tourapis, “New video quality metrics in the H.264 reference software,” Input Document to JVT, Hannover, DE, 20-25 Jul. 2008.

[5] Z. Wang, A.C. Bovik, H.R. Sheikh, and E.P. Simoncelli,“Image quality assessment: from error visibility to structural similarity,” IEEE Transactions on Image Processing, vol. 13, issue 4, pp. 600-612, Apr. 2004.

[6] Z. Wang, E.P. Simoncelli, and A.C. Bovik, “Multiscale structural similarity for image quality assessment,” Conference Record of the Thirty-Seventh Asilomar Conference on Signals, Systems and Computers, 2003, vol.2, pp. 1398- 1402, 9-12 Nov. 2003.

[7] C. Li, and A. C. Bovik, “Content-weighted video quality assessment using a three-component image model.” Journal of Electronic Imaging, vol.19, pp. 65-71, Mar. 2010.

[8] X. Ran and N. Farvardin, “A perceptually-motivated three-component image model - part I: description of the model,” IEEE Transactions on Image Processing, vol.4, no.4, pp.401-415, Apr. 1995.

[9] J. L. Li, G. Chen, and Z. R. Chi, “Image coding quality assessment using fuzzy integrals with a three-component image model,” IEEE Transactions on Fuzzy Systems, vol.12, no.1, pp. 99- 106, Feb. 2004.

[10] G. J. Sullivan and J. Ohm, “Recent developments in standardization of high efficiency video coding (HEVC),” Proc. SPIE 7798, 77980V, 2010.

[11] G. Sullivan, P. Topiwalla, and A. Luthra, “The H.264/AVC video coding standard: overview and introduction to the fidelity range extensions,” SPIE Conference on Applications of Digital Image Processing XXVII, vol. 5558, pp. 53-74, Aug. 2004.

[12] A. Puri, X. Chen, and A. Luthra, “Video coding using the H.264/MPEG-4 AVC compression standard,” Signal Processing: Image Communication, vol. 19, pp. 793-849, Oct. 2004.

[13] Z. Wang et al (2003, February), “The SSIM index for image quality assessment” [Online]. Available: https://ece.uwaterloo.ca/~z70wang/research/ssim/