VQEG Boston meeting April 2006 Motion blur : an explanation to subjective quality differences between CRT & LCD ? S. Tourancheau, S. Péchard, P. Le Callet, M. Carnec, A. Tirel, D. Barba VQEG Boston meeting April 2006

TEST SPECIFICATIONS Uncompressed HDTV (1080 50i) Motion blur : an explanation to subjective quality differencies between CRT & LCD ? TEST SPECIFICATIONS 4 CONTENTS from SVT (10s) Uncompressed HDTV (1080 50i) 7 BITRATES H.264 (JM reference encoder) main profile SAMVIQ Methodology (Random Access, explicit reference & hidden reference) Room: Rec. ITU BT 500-10 - Observation distance : 3 H MOBCAL PARKRUN 2.25 - 2.5 - 3.15 - 4 - 5 – 7 - 10 8 - 12 - 16 - 18 - 20 - 24 - 28 SHIELDS STOCKHOLM 2.25 - 3 - 4 - 5 - 6 - 7 - 8 1.63 - 1.88 - 2.25 - 3 - 3.6 - 4 - 6

DISPLAYS 1 test per display with different observers (naïve) Motion blur : an explanation to subjective quality differencies between CRT & LCD ? DISPLAYS CRT Studio Monitor JVC DTV1910CG 19’’ Interlaced scan Picture height : 20 cm LCD Philips (prototype) 37’’ (native resolution 1920x1080) Progressive scan NO POST-PROCESSINGS 1 test per display with different observers (naïve) 15 valid observers for each test

VIDEO RENDERING CHAIN SDI/DVI converter Motion blur : an explanation to subjective quality differencies between CRT & LCD ? VIDEO RENDERING CHAIN SDI/DVI converter DOREMI HDVI-20d de-interlace convert color spaces top of the art quality conv Full digital video rendering chain No lossy scaling from source to display HD PLAYER DOREMI V1-UHD - real-time playback of uncompressed HDTV SDI : 1920x1080 50i YUV 4:2:2 8bits DVI-D : 1920x1080 50p RGB 8bits CRT DISPLAY LCD DISPLAY

Motion blur : an explanation to subjective quality differencies between CRT & LCD ?

TEST RESULTS Correlation between CRT and LCD MOS : 0,938 Motion blur : an explanation to subjective quality differencies between CRT & LCD ? TEST RESULTS Correlation between CRT and LCD MOS : 0,938

Motion blur : an explanation to subjective quality differencies between CRT & LCD ? TEST RESULTS HDTV CONTENT MOBCAL PARKRUN SHIELDS STOCKHOLM MOS (CRT-LCD) 14.56 11.68 15.99 10.28 Mean IC on CRT 5.34 Mean IC on LCD 6.15

Motion blur : an explanation to subjective quality differencies between CRT & LCD ? Motion blur is induced by the type of response of LC display which is different from CRT’s one: Hold type display : LCD emitted light is sustained all the frame period (b) when CRT’s one consists in pulses (a) ; Response time : LCD response time is long with regards to classic frame rate when CRT’s one can be considered almost instantaneous.

Motion blur : an explanation to subjective quality differencies between CRT & LCD ? LCD RESPONSE TIME ISO 13406-2 : response time is the sum of rise time (from black to white) and fall time (from white to black). Manufacturers specify this response time or even only the rise time from black to white. But gray-to-gray transitions are very slowest. For the majority of displays sold before 2005, the average rise time (over gray-to-gray transitions) was superior to the frame period which leaded to annoying motion artifacts.

RESPONSE TIME COMPENSATION Motion blur : an explanation to subjective quality differencies between CRT & LCD ? RESPONSE TIME COMPENSATION Overdrive (OD) : Results (rise + fall times) : without overdrive with overdrive (Xbit Labs, december 2005)

Motion blur : an explanation to subjective quality differencies between CRT & LCD ? HOLD TYPE DISPLAY With overdrive, the rise time is less than the frame period … But even if response time was instantaneous, motion blur artifacts would be present LCD are therefore called « hold type » displays : emitted light is sustained on the screen during all the frame period … => when tracking the eyes anticipate the object motion while the image is sustained. Edges of the object are then spatially integrated on the retina during the whole frame period, resulting in a blur.

MOTION BLUR PERCEPTION Motion blur : an explanation to subjective quality differencies between CRT & LCD ? MOTION BLUR PERCEPTION CRT eye image on the retina t = 0 smooth pursuit eye movement LCD eye image on the retina

MOTION BLUR PERCEPTION Motion blur : an explanation to subjective quality differencies between CRT & LCD ? MOTION BLUR PERCEPTION CRT eye image on the retina t = T / 4 smooth pursuit eye movement LCD eye image on the retina

MOTION BLUR PERCEPTION Motion blur : an explanation to subjective quality differencies between CRT & LCD ? MOTION BLUR PERCEPTION CRT eye image on the retina t = T / 2 smooth pursuit eye movement LCD eye image on the retina

MOTION BLUR PERCEPTION Motion blur : an explanation to subjective quality differencies between CRT & LCD ? MOTION BLUR PERCEPTION CRT eye image on the retina t = 3T / 4 smooth pursuit eye movement LCD eye image on the retina

MOTION BLUR PERCEPTION Motion blur : an explanation to subjective quality differencies between CRT & LCD ? MOTION BLUR PERCEPTION CRT eye image on the retina t = T smooth pursuit eye movement LCD eye image on the retina

MOTION BLUR PERCEPTION Motion blur : an explanation to subjective quality differencies between CRT & LCD ? MOTION BLUR PERCEPTION CRT eye image on the retina t = T + T / 4 smooth pursuit eye movement LCD eye image on the retina

MOTION BLUR PERCEPTION Motion blur : an explanation to subjective quality differencies between CRT & LCD ? MOTION BLUR PERCEPTION CRT eye image on the retina t = T + T / 2 smooth pursuit eye movement LCD eye image on the retina

MOTION BLUR PERCEPTION Motion blur : an explanation to subjective quality differencies between CRT & LCD ? MOTION BLUR PERCEPTION CRT eye image on the retina t = T + 3T / 4 smooth pursuit eye movement LCD eye image on the retina

MOTION BLUR PERCEPTION Motion blur : an explanation to subjective quality differencies between CRT & LCD ? MOTION BLUR PERCEPTION CRT eye image on the retina t = 2T smooth pursuit eye movement LCD eye image on the retina

MOTION BLUR PERCEPTION Motion blur : an explanation to subjective quality differencies between CRT & LCD ? MOTION BLUR PERCEPTION CRT eye image on the retina t = 2T + T / 4 smooth pursuit eye movement LCD eye image on the retina

MOTION BLUR PERCEPTION Motion blur : an explanation to subjective quality differencies between CRT & LCD ? MOTION BLUR PERCEPTION CRT eye image on the retina t = 2T + T / 2 smooth pursuit eye movement LCD eye image on the retina

MOTION BLUR PERCEPTION Motion blur : an explanation to subjective quality differencies between CRT & LCD ? MOTION BLUR PERCEPTION CRT eye image on the retina t = 2T + 3T / 4 smooth pursuit eye movement LCD eye image on the retina

Motion blur : an explanation to subjective quality differencies between CRT & LCD ? BLUR WIDTH PERCEPTION We set an experiment in order to measure the blur width of a moving edge, changing its velocity. The stimuli used was a periodic structure of vertical (resp. horizontal) bars moving from left to right (resp. from down to up) on a black background. Displayed and perceived stimuli :

EXPERIMENT : TASK The scrolling of the bars was continuous. Motion blur : an explanation to subjective quality differencies between CRT & LCD ? EXPERIMENT : TASK The scrolling of the bars was continuous. Using the arrow keys of a keyboard, the observer can increase or decrease, in real-time, the space between the bars. He can operate as many times as he wants, until he considers that the two blurred areas are just merging. When he is satisfied, he validates his measure and then the next presentation is displayed.

EXPERIMENT : PROTOCOL 7 observers : Protocol : Motion blur : an explanation to subjective quality differencies between CRT & LCD ? EXPERIMENT : PROTOCOL 7 observers : all of them were familiar with the procedure and have a perfectly corrected sight. Protocol : 1 session consists in a set of 17 presentations, concerning four types of stimuli (vertical/horizontal and white/red) for different velocities, the sequence is totally random, each observers repeated the test twice, on a different day average lenght of a session was between 10 and 15 minutes.

RESULTS Mean Blur Width Curve fitting give us the following model : Motion blur : an explanation to subjective quality differencies between CRT & LCD ? RESULTS Mean Blur Width Curve fitting give us the following model : with where : w is the blur width in pixels v is the velocity in pel / seconds T is the frame period V = vT is the velocity in pel / frames

Motion blur : an explanation to subjective quality differencies between CRT & LCD ? RESULTS AND MODEL Results and model for a white stimulus moving with an horizontal movement :

Motion blur : an explanation to subjective quality differencies between CRT & LCD ? PAN’s MODEL Pan, H., Feng, X.-F. et Daly, S. (2005). LCD motion blur modeling and analysis. IEEE International Conference on Image Processing, II:21–24. Mathematical model to predict the blur width in function of velocity and in which the temporal response of LCD is a parameter. They studied for examples two types of response, linear and sinusoidal one, the model give the following results : As we can see, the sinusoidal response gives model wich are very close to our results. ( 1.047(vT) )

Motion blur : an explanation to subjective quality differencies between CRT & LCD ? PAN’s MODEL When a ideal instant temporal response (a) of LCD is used, a blur width of 0.8vT is obtained, only 25% motion blur is caused by slow response on a standard LCD panel. Hold-type is responsible for 75% of motion blur. In the same way, reduction methods can be evaluated by modifying the temporal response of LCD : Black data insertion (b) Back light flashing (c) Frame rate doubling method and motion compensated inverse filtering

Motion blur : an explanation to subjective quality differencies between CRT & LCD ?

Motion blur : an explanation to subjective quality differencies between CRT & LCD ?

Motion blur : an explanation to subjective quality differencies between CRT & LCD ?

Motion blur : an explanation to subjective quality differencies between CRT & LCD ?