The Effects of Compressed Acquisition on Downstream Video Quality Discovery Communications Discovery Development and Technology Group.
Post on 23-Dec-2015
216 Views
Preview:
Transcript
The Effects of Compressed Acquisition on Downstream Video QualityDiscovery CommunicationsDiscovery Development and Technology Group
Overview
• File based acquisition • Acquisition file formats and
compression• The post and distribution chain• Test procedures• Results• Conclusions
The HD Acquisition Ecosystem
HDCAM
HDCAM SR
HDV
DVCPRO HD
P2
XDCAM
INFINITY
RED
XDCAM(Flat Tape)
Firestore(Field Drives)
Tape Based Systems
File Based Systems
Commonalities of file based cameras
• Use flash memory or hard drives– Flash based media increasing in capacity,
decreasing in cost
• Often use new codec technology– Compression efficiency increasing, new
codecs
• Enable tapeless workflows in post– Support for new workflows and codecs
continues to grow
Today’s Compression Schemes in Post and Acquisition
Bit Rate(mb/s)
Inter/Intra
GOPLength
Down Sampling Chroma Sampling
HDCAM SR 480 Intra I None 4:2:2
HDCam 140 Intra I 1440x1080 3:1:1
DVC Pro 100 Intra I 1280x1080 2.6:1.3:1.3
J2K (Thomson)
100,75,50 Intra I Variable Variable
AVC Intra 100, 50 Intra I1440 by 1080 on
50 mbps4:2:2, 4:2:0 at 50
mbps
XDCAM HD35, 25,
18 LGOP 15 1440 x 1080 4:2:0
HDV 25 LGOP 15 1440 x 1080 4:2:0
VC-3 (DnX) 36, 145,
220 Intra I None 4:2:2
Apple Pro Res 145, 220 Intra I None 4:2:2
A long chain to the consumer
Acquisition Compression
NLE / Post Compression
Tape Delivery /
Compression
Network Post Compression
Air Master Compression
Playout Server Compression
Broadcast Compression
MSO/ Distributor
Compression
HDCAM
XDCAM
DVCPRO
AVC Intra
JPEG 2000
HDV
MPEG Intra
DnX
Pro Res
AVC Intra
DVCPRO
HDCAM SR
HDCAM
D5
DVCPRO
DnX
Pro Res
AVC Intra
DVCPRO
HDCAM SR
HDCAM
D5
DVCPRO
MPEG-2
AVC Intra
JPEG 2000
MPEG-2
AVCMPEG-2
AVC
Test Scope
• Objective: – Discover the effects of different codecs and bit
rates in the acquisition, post production, and transmission processes on video quality delivered to the home consumer.
• Test Devices– Sony F23: Digital Cinema camera, records to
HDCAM SR– Panasonic HPX-3000: Broadcast camera, records
to P2 cards using AVC Intra or DVCPRO codecs– Thomson Infinity: Broadcast camcorder, records to
Rev Pro discs using JPEG 2000 or MPEG-2 I frame compression
– P2 Mobile: Panasonic field recorder for P2• Test results are not the final word on these
codecs, but a reflection of their implementation in these devices.
Test Procedures – Camera Tests
• Discovery has a policy of testing cameras before they are allowed to be used on our productions– Studio Tests
• Standard studio battery with sets designed to challenge cameras
– Field Tests• Location shoots, often outdoors, with optically challenging
material
• All footage used in these tests was gathered during our regularly scheduled camera testing.
Test Procedures – Post Tests
• Tests designed to assess the effect of mezzanine level post production compression on various codecs– Materials created in the camera codecs were
encoded to the following mezzanine post production codecs:
• Avid DnX (VC-3) at 145 mb/s• Avid DnX (VC-3) at 220 mb/s
– Materials were output to both HDCAM and HDCAM SR for transmission testing to determine the impact of additional tape format compression.
• HDCAM: 145 mb/s, DCT-based compression• HDCAM SR: 480 mb/s, MPEG-4 Intra Frame compression
Test overview – Transmission Tests
• Transmission tests were performed using backup playout and air chains. Four different transmission scenarios were tested:– MPEG-2 transmission with air chain devices – MPEG-2 transmission without air chain– MPEG-4 transmission with air chain devices – MPEG-4 transmission without air chain
MPEG-2 Air Chain
MOTOROLADIGICIPHER-2
MULTIPLEXER
MOTOROLADIGICIPHER-2
FEC/PHASE MODULATOR
SATELLITE SYSTEMS5420
L-BANDUPCONVERTER
MOTOROLADSR-4410
INTEGRATED RECEIVER/DECODER
MOTOROLA HDD-1100
HD DECODER
MOTOROLADIGICIPHER-2
HIGH DEFINITIONENCODER
DS-3 DS-3 70-MHz IF L-BAND
L-BAND ASIHD-SDI
INGEST VTRHDW-5500
OMNEON SPECTRUM
SERVERCAPTURE MPEG-2 AT
60 Mb/s 4:2:2
MIRANDA INTUITION
IPM AND BRANDING
HDSDI
HDSDI HDSDI
RECORD VTRHDW-5500
MPEG-4 Air Chain
MOTOROLATMX-2010
ASI MULTIPLEXER
MOTOROLASEM DCII
SATELLITEENCRYPTER
NEWTEK
DVB-S2MODULATOR
MOTOROLADSR-4410MD
MULTI-DECRYPTER
SENCOREMRD-3187A
HD DECODER
MOTOROLASE-5100A
MPEG-4/AVCENCODER
ASI ASI ASI 70 MHz IF
L-BAND ASI
SATELLITE SYSTEMS5420
L-BAND UPCONVERTER
70 MHz IF
INGEST VTRHDW-5500
OMNEON SPECTRUM
SERVERCAPTURE MPEG-2 AT
60 Mb/s 4:2:2
MIRANDA INTUITION
IPM AND BRANDING
HDSDI HDSDI
RECORD VTRHDW-5500
HDSDI
Reviewing the results
• Test results were reviewed using three different methods:– Double blind viewer testing using ITU 500
criteria for expert viewing. – PQA Analysis using the Tektronix PQA
500. Tests were run modeling both expert viewers and typical viewers. Tests included PSNR, PQR, and DMOS models.
– Split screen comparison of test results and codecs.
Results – Visual Analysis Acquisition Codecs
• The visual effects of the different acquisition compression formats were apparent at the pre-transmission stage to our expert viewers.– AVC Intra at 100 mb/s had the least noticeable
impairments, performing well across all the image tests.
– JPEG 2000 at 100 mb/s had similar performance as AVC in low complexity images, but there were noticeable increase in visual impairments in more complex images.
– AVC Intra at 50 mb/s performed well in some tests, albeit with a predictable impairment due to reduced bit rate.
– JPEG 2000 at 50 mb/s had a high level of visible artifacts with most of our test images.
Results – Visual Analysis Post Production Path
• The different compression schemes typically made minor differences in the post production chain.
- Expert viewer panels could see concatenation effects when applied to the uncompressed source.
- Effects were less obvious with AVC and J2K at 100 mb/s.
- As bit rates were reduced the effects became more pronounced and easily identified by the test panel.
Results – Visual Analysis Transmission Path
• MPEG-2 with Air Chain– AVC scored consistently, with the 100 mb/s
clips occasionally being confused with the master uncompressed clip by our expert viewers
– JPEG 2000 scored far less consistently with both 100 and 50 mb/s clips being scored randomly throughout the range, particularly by the typical viewer panel. The expert viewer panel tended to always score J2K 50 mb/s as having the highest level of visual impairment.
Results – Visual Analysis Transmission Path
• MPEG-2 without Air Chain– The master clip was clearly recognized by our
expert viewers in this test, scoring the highest– AVC Intra 100 and 50 scored comparatively
well, and had results slightly below the master clip’s.
– JPEG 2000 scored lower, with 100 megabits typically in the middle of the results and 50 megabits almost always at the bottom of the rankings
Results – Visual Analysis Transmission Path
• MPEG-4 AVC Tests– Both viewer panels had a much harder time
judging the MPEG-4 transmission tests, both with and without the air chain.
• The typical viewer panels were usually able to distinguish differences, but frequently unable to determine what constituted a higher quality image.
• The expert viewer panel consistently rated the uncompressed master at the top with both AVC and the J2K 100 clips following closely. J2K at 50 mb/s was rated at the bottom of both MPEG-4 tests.
– The level of total picture degradation and the softening of the image made it harder to see differences in the picture. The picture softness may be caused by our current AVC decoder.
Results - PQA
• PQA results were less conclusive.– The PQA typical viewer models produced
a variety differences in the results. • In simpler sequences, JPEG2000 was often
the leader• In more complex sequences AVC Intra was
often the forerunner.
– Expert viewing models demonstrated more predictable results.
Results - PQA
• For the MPEG-2 and MPEG-4 transmission paths the PQA tended to correlate with the viewer test results.– MPEG-4 results showed little difference
between the codec concatenations with or without the air chain devices.
– They also showed much lower scores for the sequences overall as compared to the MPEG-2 results.
– As previously mentioned, the MPEG-4 decoder may have contributed to these results.
Conclusions
• Acquisition compression formats are not necessarily damaging to the final image quality when used at higher bit rates.
– At the higher bit rates, expert viewers sometimes scored the acquisition codecs equal to or higher than the uncompressed master after MPEG-2 transmission.
– Lower bit rate codecs were clearly distinguishable at any stage of codec concatenation throughout the acquisition, post, and transmission paths.
Conclusions
• Compression applied in the transmission paths can be a “blender” tending to mask initial visual differences.– There were minimal noticeable differences with
the air chain devices.– Clear differences between the MPEG-2 and
MPEG-4 compression processes in all tests.
• Visual impairments induced by codec concatenation in the production and transmission process are generally not discernable to the average viewer, and in many cases are statistically irrelevant even to expert viewers at critical viewing distances.
Follow up for more answers and details
The Discovery Development and Technology Group
Bert CollinsJosh DerbyCharlie Myers
E-mail us at ddtg@discovery.com
top related