BBT - Functional Testing of Multimedia Devices

Picture Quality Metrics

  • Blocking effect - result of quantization
  • Ringing effect - high frequency noise near edges
  • Blurring effect - blurring due limited bit-rate
  • Field loss - "interlace" like artifacts

Various Supported Platforms

  • IBM Cell BE processor (1+8 cores @3.2 GHz, SIMD)
  • Intel Multi-core platform (Intel Mobile Core2Quad processor Q9000 series, 4 cores@ 2.0 GHz, SSE 4.1)
  • FPGA (Xilinx Virtex IV XC4VLX100-10)
  • introduced by PSF offset

System Modularity

  • Customizable set of artifacts in delivery package
  • Measures are independent from each other
  • Every measurement algorithm is
  • implemented as separated module

Results Interpretation

Measures produce frame-based subjective score regarding effect being monitored

 

Optimization

Optimized using target architecture parallel potential (multi core, vector instruction set, FPGA platform)

Picture Quality Metrics

RT-IP030

pqm_semaphoreRT-IP030 or Picture Quality Meter (PQM) is a no-reference realtime processing system for video quality assessment in broadcast systems and multimedia. It provides real-time monitoring, warning and tracking of measured results of video sequences.

PQM addresses artifacts coming from MPEG-like compression schemes and broadcasting process. PQM gives frame-based output and can handle up to HD 1080i@60 live sources. 

RT-IP030 Technical Data

PQM Metrics

  • Blocking effect - result of quantization
  • Ringing effect - high frequency noise near edges
  • Blurring effect - blurring due limited bit-rate
  • Field loss - "interlace" like artifacts introduced by PsF offset

System Modularity

  • Customizable set of artifacts in delivery package
  • Measures are independent from each other
  • Every measurement algorithm is implemented as separate module
  • Measures produce frame-based subjective score regarding effect being monitored
  • Optimized using target architecture parallel potential (multi-core, vector instruction set, FPGA platform)

Various Supported Platforms

IBM BladeCenter solution

  • Linux platform
  • Cell SDK
  • 1Gbps Ethernet interface
  • 1080i60 streaming over LAN
  • lossless compression inside grabber device
  • remote monitoring (PQM Monitor)
  • proven in real-time on IBM Cell BE processor (1 + 8 cores @ 3.2 GHz, SIMD)

PQM embedded solution – Intel Multi-core platform

  • Supports commercially available grabber card (SD/HD)
  • OpenMP, OpenCL framework
  • DirectShow libraries
  • proven in real-time on Intel Mobile Core2Quad processor Q9000 series (4 cores @ 2.0 GHz, SSE 4.1)

FPGA RTL only solution

  • RTL only solution for artifact detection
  • universal video input interface supports all data rates up to 1080p60 (LVDS, HDMI, CVBS, etc)
  • real-time video output with integrated visualization of results (optional)
  • measurement results visible in I2C register map (optional)
  • proven in real-time on Xilinx Virtex IV FPGA running at 150MHz with 1080p60 input video stream (lower video resolutions are tested as well)

RT-IP030 Artifact Examples

pqm_blocking1s

Blocking

  • Detects the compression grid (usually 8x8)
  • Also handles scaled grid
  • Measures the differences across grid
  • Intra-frame algorithm
  • Gives an overall quality score for one frame (good, unreliable, bad)
pqm_blocking2s
pqm_ringing1s

Ringing

  • Detects potential areas of ringing (strong edges)
  • Analyzes the area around strong edges to distinguish between texture and ringing
  • Measures the ringing amplitude and combines it with edge strength for final figure
  • Intra-frame algorithm
  • Gives an overall quality score for one frame (good, unreliable, bad)
pqm_ringing2s
pqm_blurring1s

Blurring

  • Determines the sharpness of the frame
  • Calculates overall blurring by inverting sharpness value
  • Intra-frame algorithm
  • Gives an overall quality score for one frame (good, unreliable, bad)
pqm_blurring2s
pqm_fod1s

Field Loss

  • Determines if two consecutive frames belong to same instant of time (same original frame, due to interlacing process)
  • Process only progressive input
  • Artifacts evaluated only in area where motion is detected
  • Inter-frame algorithm
  • Gives an overall quality score for one frame (artifacts present, not detected)
pqm_fod2s

RT-IP030 Use Cases

  • Video quality assessment of the broadcasting signal
  • Quality assessment of multi-media devices
  • DTV back-end quality assessment
  • STB video output quality assessment
  • Detection of the up-scaled SD material to HD (blurring measurement)
  • Detection of the bad deinterlacer temporal processing (field offset detection)
pqm_use_s

RT-IP030 Resource Utilization


Blocking Ringing Blurring FOD    All  
IBM Cell BE
(1+8 cores @ 3.2 GHz, SIMD)		 Cores 2 3 1 1 7
Memory 8MB 10MB 8MB 4MB
Intel Q9000
(4 cores @ 2.0 GHz, SSE 4.1)		 Cores 3 4 1 1 9
Memory 6MB 8MB 6MB 4MB

Blocking Ringing Blurring FOD    PQM_FPGA_IP
Xilinx Virtex IV
XC4VLX100-10
 Clock 155MHz
 156MHz 173MHz 185MHz 154MHz
BlockRAM 88 42
 9 1 229
DSP 11 33
 12
 0
 62
SLICE 7098 13741 2854 273 40542
Altera Stratix II
EP2S90-3
 Clock 188MHz 187MHz 212MHz 232MHz 173MHz
Memory [bits] 1217464 651264 101376 32768 1839032
DSP 22 24 0 0 46
ALUT 16475 32267 7667 870 59502
External Line Buffers
 3
 32
 6
 0
 32 line buffers are embedded
in PQM_FPGA_IP module
External Memory Frame Size:
1920x1080
 Data width:
16 bits
 -
 -
 Data width:
  8bits		 Two frames:
1 x 16 bits per frame
1 x 8 bits per frame

IBM CELL BE INTEL MC FPGA (RTL)
Algorithm libraries
 Binary libraries, Linux
 Binary libraries, WinXP NET list
Demo application in source code Yes, ANSI C Yes, MS VS2005 Yes, FPGA wrapper
SNMP interface No Yes Yes
Documentation Yes Yes Yes
Test sequences Yes Yes Yes
Demo Platform* RT-AV101 + PS3 RT-AV101 + PC RT-AV200
Integration Support* Yes Yes Yes
*Not included in IP package by default, it is optional and charged separately.