The Lecture Contains: Frequency Response of the Human Visual System: Temporal Vision: Consequences of persistence of vision: file:///d /...se%20(ganesh%20rana)/my%20course_ganesh%20rana/prof.%20sumana%20gupta/final%20dvsp/lecture8/8_1.htm[12/31/2015 11:11:04 AM]
Frequency Response of the Human Visual System: Any video system is targeted for human viewers. So it is important to understand how HVS perceives video signals. We will focus on perception of spatial & temporal changes in image luminance. Similar results hold for perception of chrominance variation, being proportionally lower (less) than for the luminance. It will be shown that visual sensitivity to a visual pattern depends on the pattern's spatial and temporal frequency content. The visual sensitivity is highest at some intermediate spatial & temporal frequencies. It then falls off quickly & diminishes at some cut off frequencies. Spatial or temporal changes above these frequencies are invisible to the human eye. Knowledge of the visual frequency response is very important in designing a video system. For example, the temporal & spatial cut off frequencies form the basis for determining the frame rate & line rates in a video capture and display system. We first describe the temporal frequency response of HVS. Temporal Vision: Temporal masking and perception of temporally changing stimuli are extremely important in interframe coding. However, temporal masking is complicated by at least two facts: 1. Television cameras integrate the image of any object on the target and, thus, there is motionrelated blurring and resolution loss; 2. Perception of a moving object depends heavily on whether or not the object is tracked by the eye. file:///d /...se%20(ganesh%20rana)/my%20course_ganesh%20rana/prof.%20sumana%20gupta/final%20dvsp/lecture8/8_2.htm[12/31/2015 11:11:04 AM]
The psychophysical literature contains many facts about the perception of temporally changing stimuli. However, their application to coding is still in its infancy. Instead, several applied studies have attempted to evaluate the loss of perceived resolution (spatial and amplitude) as a result of movement in scenes. If movement is drastic, such as with a scene change (when TV cameras are switched), the perceived spatial resolution is reduced significantly immediately after the scene change. In fact, the perceived spatial resolution of the new scene may be reduced down to only onetenth of normal without detection provided that full resolution is restored gradually within about half a second. Experiments also show that if a moving object is tracked by the eye, then perceived resolution due to camera integration dominates any reduction in resolution introduced by the visual system. However, when erratically moving objects are not tracked, the loss of perceived spatial resolution due to the visual system is significant. In practical television viewing, most displayed movement is not easily tracked. However, we have no quantitative data to tell when a viewer tracks an object and how accurately he tracks it. Also, since in many visual communication systems a transmitted picture may be viewed by many observers (e.g. broadcast TV), it is not clear how the resolution loss of nontracked objects can be used to improve the coding efficiency. The temporal frequency response of HVS refers to the visual sensitivity to a temporally varying pattern at different frequencies. Experiments are performed to determine the temporal frequency response of the HVS. Note: it is found that the temporal response of an observer depends on many factors, including, 1. Viewing distance, 2. Display brightness,and 3. Ambient lighting. file:///d /...se%20(ganesh%20rana)/my%20course_ganesh%20rana/prof.%20sumana%20gupta/final%20dvsp/lecture8/8_3.htm[12/31/2015 11:11:05 AM]
The experiment performed by Kelly[ ] was as follows: The viewer was presented a flat' screen whose brightness varied sinusoidally as: In other words the pattern is spatially constant and varies only temporally. For fixed mean brightness B & frequency f, the modulation level m' was varied and the viewer was asked to identify the lowest modulation level at which the temporal variation of the screen brightness (i.e. flicker) became just noticeable. The inverse of represents how sensitive the eye or viewer is to the temporal changes at the given frequency. The factor is referred to as contrast sensitivity. This is used to describe the visual sensitivity or response. The contrast sensitivity as a function of frequency is referred to as the Modulation Transfer function of HVS. It can be seen that the temporal response of HVS is similar to that of a band pass filter, with peaks at some intermediate frequency and the response falls off quickly afterwards, up to 4 to 5 times the peak frequency. The peak frequency increases with mean brightness of image. This is shown in figure below. Both the average brightness ( luminance) and modulation amplitude are expressed in trolands which is a unit of retinal illuminance. A troland is defined as the retinal illuminance when a surface of illuminance is viewed through a pupil at the eye of area. As a rough approximation, for display viewing 100 trolands is equivalent to about in the display. As an example, at a mean brightness of 0.65 trolands, the peak response occurs at cut-off frequency where the response essentially diminishes, is at about 20-25Hz. and the On the other hand at mean brightness of 850 trolands, the response is highest at about 15-20 Hz and diminishes at about 75Hz. One reason why eye has reduced sensitivity at higher temporal frequency is because the eye can retain the sensation of an image for a short time interval (persistence period) even when the actual image has been removed. This is referred to as the phenomena of persistence of vision. file:///d /...se%20(ganesh%20rana)/my%20course_ganesh%20rana/prof.%20sumana%20gupta/final%20dvsp/lecture8/8_4.htm[12/31/2015 11:11:05 AM]
Consequences of persistence of vision: 1. On one hand, this causes temporal (blurring of the observed pattern, if a pattern changes at a rate faster than the refresh rate (i.e., of HVS. The visual response falls off very quickly beyond this frequency. Note: Persistence period. The brighter the light source, shorter the persistence period, or higher the refresh rate of HVS. This implies higher temporal sensitivity or large temporal cut off frequency for large brightness. 2. On the other hand, this vision persistence allows the display of a video signal as a consecutive sequence of frames. This implies, if (frame interval) persistence period, eye perceives continuously varying image; whereas If persistence period, the eye will observe frame flicker which refers to perception of discontinuous frames. The lowest frame rate at which eye does not perceive flicker is known as the critical flicker frequency By definition, this is equivalent to the temporal cut-off frequency. The frame rate used by a video capture or display system should be greater than critical flicker frequency to avoid perception of flicker. file:///d /...se%20(ganesh%20rana)/my%20course_ganesh%20rana/prof.%20sumana%20gupta/final%20dvsp/lecture8/8_5.htm[12/31/2015 11:11:05 AM]
From the curves shown in figure ( ), we observe 1. That the critical flicker frequency ranges from 20 to 80Hz, depending on the mean brightness of display. 2. The brighter the display, the higher the critical frequency. That is, the eye is noticeably more sensitive to flicker at high luminances than at low luminances. 3. The 9300 troland curve is appropriate for the heights of the bright CRT displays, for which the flicker sensitivity peaks at about 15Hz. Cut-off is reached at 70Hz and there is a substantial fall off at low frequencies. 4. The 0.65 troland curve, appropriate for low light displays, shows a much reduced sensitivity to flicker, lower bandwidth and much less fall off at low frequencies. This is why motion picture system uses a lower frame rate than TV system. Presently the motion picture industry uses 24 frames/sec, whereas TV industry uses 50 (fields/sec)(pal) or 60 (fields/sec) (NTSC). We note that these frame and field rates are close to the cut-off frequency associated with the mean brightness levels in their intended applications. Computer displays use a much higher rate i.e., 72 Hz(fps). This is because a computer user sits closer to the screen than does a TV viewer, and at a shorter distance the visual threshold is higher. Other factors affect the flicker sensitivity as well. The flicker sensitivity increases gradually with increasing stimulus area. It is less for peripheral vision than for foveal vision. file:///d /...se%20(ganesh%20rana)/my%20course_ganesh%20rana/prof.%20sumana%20gupta/final%20dvsp/lecture8/8_6.htm[12/31/2015 11:11:05 AM]