Title: Animation Speed
1Animation Speed
- Question What should be done when the processes
of scan, convert, erase redraw takes longer
than 50 msec?
Display
Frame Buffer
Scan-converting process
Erase redraw
time
40 msec
gt10 msec
2Use Double Buffering
- Frame buffer is divided into two images. Each
with half bits per pixel of the overall frame
buffer
1st half
Erase redraw
Scan-converting
2nd half
Scan-converting
Erase redraw
time
25 msec
25 msec
25 msec
3Course Progress
- Introduction
- Data Representation
- Processing Techniques
4Data Representation
- Number and text
- Graphics and animations
- Sound and Audio
- Images Representation
- Video Representation
5Sound and Audio
- Sound Concept How is analog sound represented in
digital data? - Human Auditory System
- Digitization of sound waves
- Standard Sound and Audio formats
6Sound and Audio
- Sound wave is a longitudinal wave of air
pressure. - Unlike Electromagnetic waves, the change in air
pressure is in the same direction as the
propagation of the sound.
time
7Sound and Audio
- An audio sound wave is characterised by the pitch
and loudness. - Sound wave can be represented by a waveform with
frequency and amplitude s.t. - Frequency measures pitch
- Amplitude measures loudness
time
8Amplitude
- Dynamic Range is the difference between the
quietest and the loudest sound. - Amplitude measures the loudness of the sound
Amplitude
time
0 1 2 3 4 5 6
7 8 9 10 11 12
9Amplitude
- Amplitude is measured in units of deciBel (dB).
- Low background noise
- Normal speaking
- High uncomfortable to ears (gt80dB)
- X units in Y dB means
- Y 20 log10X
Amplitude
time
0 1 2 3 4 5 6
7 8 9 10 11 12
10Wavelength
- Air pressure of sound waves oscillates in cycles.
- Wavelength is the length of a cycle period of a
waveform. - Wavelengths can be measured in seconds or
milliseconds.
wavelength
time
0 1 2 3 4 5 6
7 8 9 10 11 12
11Frequency
- Frequency 1/wavelength
- Frequency in Hertz (Hz) 1Hz 1/second
- Infra sound 0-20 Hz
- Human hearing 20Hz - 20KHz
- Ultrasound 20KHz 1 GHz
- Hypersound gt1GHz
wavelength
time
0 1 2 3 4 5 6
7 8 9 10 11 12
12Sound and Audio
- A pure tone is a sine wave.
- Two waves with the same frequency may start at
different time. Phase or phase shift refers to
the relative time delay between two waves.
time
phase
13Sound and Audio
- Sound waves are additive. In general, the
resultant sound wave is represented by a sum of
sine waves.
time
Component tones
Resultant sound wave
14Human Auditory System
- Human ear converts sound waves to nerve signals
- Brain neurons recognize the nerve signals
- Human ears can hear
- Frequency range of 20Hz to 20 KHz.
- Dynamic range of 120 dB.
- Ears are uncomfortable above 80dB.
15Human Auditory System
- Psychoacoustic studies show that we cannot hear
some sound for some psychoacoustics principles - Absolute threshold of hearing
- Critical band frequency
- Simultaneous masking
- Temporal masking
16Absolute Threshold
- Absolute threshold of hearing is the amount of
energy needed in a pure tone - This threshold depends on the frequency
- The absolute threshold can be approximated by
- in dB SPL, where SPL is the sound pressure
level.
17Absolute Threshold of Hearing
18Absolute Threshold
- The lowest point on the curve at around 4kHz.
- Tq(f) may be interpreted as a maximum allowable
energy level for coding distortions introduced in
the frequency domain.
19Critical Band Frequency
- Frequency-to-place transformation takes place in
the inner ear along the basilar membrane - Each region in the cochlea is attached with a set
of neural receptors - Distinct regions in the cochlea are tuned to
different frequency bands - Critical bandwidth can be loosely defined as the
bandwidth at which subjective responses change
abruptly
20Critical Band Frequency
- The detection threshold for a narrow-band noise
source between two masking tones remains constant
as long as the frequency separation between the
tones remains within a critical bandwidth. - Beyond the bandwidth, the threshold rapidly
decreases.
21Critical Band Frequency
Sound Pressure Level (dB)
Audibility Threshold (dB)
?f
masking sources
Freq.
?f
Critical bandwidth
measured sources
22Simultaneous Masking
- Simultaneous Masking (SM)
- one sound is rendered inaudible because of the
presence of another sound - Presence of a strong masker creates an excitation
of sufficient strength on the basilar membrane at
the critical band location to block transmission
of a weaker signal effectively - Spread of Masking
- A masker centered with one critical band may
affect the detection thresholds in other critical
bands.
23Simultaneous Masking
- 2 types of simultaneous masking depending on the
relative strength of the tone and noise - Tone-masking noise
- Noise-masking tone
- Noise-masking threshold is given by
- THN ET 14.5 B
- where ET is energy level of critical band
tone-masker and B is the critical band number. -
24Simultaneous Masking
- Tone-masking threshold is given by
- THT EN K
- where EN is energy level of critical band
noise-masker and K is typically between 3 and 5
dB. - Masking thresholds are functions of Just
Noticeable Distortion (JND). -
25Temporal Masking
- Absolute audibility thresholds for masked sounds
are increased prior to, during, and following the
occurrence of a masking signal. - A listener will not perceive signals beneath the
elevated audibility thresholds produced by abrupt
signal transients. - Thus, abrupt signal transients create pre-masking
and post-masking regions for a time period.
26Temporal Masking
Masked Audibility Threshold Increase
Simultaneous masking
Post-masking
Pre-masking
50
100
0
50
150
200
-50
0
Time after masker appear(ms)
Time after masker removal (ms)
27Temporal Masking
- Pre-masking tends to last only about 5ms.
- Post-masking will extend from 50 to 300ms,
depending upon the strength and duration of the
masker. - Temporal masking has been used in audio-coding
algorithms. - Pre-masking has been exploited in conjunction
with adaptive block-size transform coding to
compensate for pre-echo distortions.
28Sound Representation
- The analog sound wave is a one dimensional wave
of air pressure. - Mathematically, the digital sound wave is
represented as a one dimensional digital waveform
in computers. - The amplitude of the waveform is the measured air
pressure of the sound wave.
time
0 1 2 3 4 5 6
7 8 9 10 11 12
29Digitization of Sound
- A microphone receives the sound waves and
converts the sound waves into electronic signals
in the analog waveforms. - An A/D converter performs the analog-to-digital
signal conversion to changes the analog signals
into digital signals. - Computers can store, transmit, and process the
digital signals. - A D/A converter performs the digital-to-analog
signal conversion. - The speakers converts the electronic signals into
sound waves.
30Digitization of Sound
A/D Converter
1001
D/A Converter
1001
31Digitization of Sound Wave
- A/D converter takes samples of the amplitude of
the analog signals at different time positions. - The number of samples per second, called the
sampling rate, is fixed.
time
0 1 2 3 4 5 6
7 8 9 10 11 12
Time for 1 sample
32Amplitude
- The sample values can be encoded with more or
less bits. More bits can describe the amplitude
in finer details. - 8 bit quality ? 256 different values
- 16-bit quality ? 65,536 different values
- The sampled values are then encoded in a format,
such as Pulse Code Modulation. PCM uses fixed
length for each frequency sample.
33Encoding Formats
- Pulse Code Modulation (PCM) digital
representation of analog signal where the
magnitude of the signal is sampled regularly at
uniform intervals and then quantized to a series
of symbols in a binary code - Differential PCM record the differences between
samples instead of the sample values - Adaptive Differential PCM record differences
between samples and adjust the coding scales
dynamically.
34Standard Audio/Sound Formats
- Telephone quality speech input
- Mono-channel
- 8,000 samples/second, 8 bits/sample
- CD quality stereo audio
- 2 stereo channels (left right)
- 44,100 samples/second, 16 bits /sample
35Standard Audio/Sound Formats
- DVD quality audio
- Mono(1.0), stereo(2.0), Quad(4.0) or
surround(5.1) channels - 5.1 channels Left, Right, Centre, Left Rear,
Right Rear, subwoofer - 16, 20, or 24 bits per sample
- 44.1kHz, 48kHz, 88.2kHz, or 96kHz
- PCM format
- Max bit rate 9.6 Mbps
36Standard Audio/Sound Formats
- MIDI (Musical Instrument Digital Interface)
- Digital encoding of musical information
- The sound data is not used. Only the commands,
music scores, that describe how the music should
be played are used. - Highest compression, easy for editing
- Requires a music synthesizer to generate music
37Course Progress
- Introduction
- Data Representation
- Number and text
- Graphics and animations
- Sound and Audio
- Images Representation
- Video Representation
- Processing Techniques
38Course Progress
- Image Representation
- Light and Colour
- Human vision system
- Image capture
- Image quality measurements
- Image resolution
- Connectivity and distance
- Colour representation models
- Camera calibration
39Light and Colour
- Light
- Electromagnetic Wave
- Wavelength 380 to 780 nanometers (nm)
- Colour
- Depend on spectral content (wavelength
composition) - E.g. Energy concentrated near 700 nm appears red.
- Spectral colour is light with a very narrow
bandwidth. - A white light is achromatic.
40Colour
- Two types of light sources
- Illuminating light source
- Reflective light source
41Colour
- Illuminating light source generates
- Light of certain wavelength, or
- Light of a wide range of wavelength
- Follow additive rule
- Colour of the mixed light depends on the SUM of
the spectra of all the light sources
42Colour
- Examples of illuminating light sources of wide
wavelengths - Sun, stars
- Light bulb, florescent tube
- Examples of illuminating light sources of narrow
wavelength - Halogen lamp, Light power light bulb
- Phospher, Light Emitting Diode
43Colour
- Reflecting light
- Object absorbs incident light of some wavelengths
- Object reflects incident light of remaining
wavelengths - E.g. An object that absorbs wavelengths other
than red would appear red in colour - Follow subtractive rule colour of the mixed
reflecting light sources depends on the remaining
unabsorbed wavelengths
44Colour
- Examples of reflecting light source
- Mirror and white objects reflects all wavelengths
of light regularly - Most solid objects
- E.g. Dye, photos, the moon
45Colour
- Complements of colours
- Red ? Cyan
- Green ? Magenta
- Blue ? Yellow
46Human Vision System
- Schematic diagram of a human eye
Circular muscle
Lens
Fovea
Optic nerve
Cornea
Retina
Iris
47Human Vision System
- Each part of our eye has some functions
- Iris controls the intensity of light entering the
eye - Circular muscle controls the thickness of the
lens - Lens refracts the light onto the fovea
- Optic nerve transmits light signals to the brain
- Brain
- interprets the light signals from both eyes
- understands the signals as an image
48Human Vision System
- Path of light in the human eye
- Enters eye through cornea
- Passes through hole within iris
- Refracted by the lens
- Hits the retina wall inside the eye
- Excites some light-sensitive cells
49Human Vision System
- Path of visible light inside a human eye
Circular muscle
Lens
Fovea
Optic nerve
Cornea
Retina
Iris
50Human Vision System
- 2 types of light-sensitive cells
- Cones and rods due to their shapes
- Rods
- Very sensitive to intensity of light
- Generates monochromatic response
- More sensitive in low light level than cones
- Distributed across the retina wall
- See grey images under dim light at night
51Human Vision System
- Cones
- red cone, green cone, and blue cone
- Each type of cone is sensitive to light of
different wavelength - Densely populated at the fovea. Approx. 160,000
cones per mm2 - Size of cones 1µm across an angle of 20 seconds
of arc - Smallest readable object 0.2mm at 1m far
52Human Vision System
- We can distinguish only 40 shades of brightness
- Our brightness sense adapts as we scan over an
image - Most colours could be matched by mixing different
portions of a red, a green, and a blue primary
source.
53Image Capture
- Images may be captured using
- Cameras
- Video cameras
- Fax machines
- Ultrasound scanners
- Radio telescopes
- An image is formed by the capture of radiant
energy that has been reflected from surfaces that
are viewed
54Image Capture
- The amount of reflected energy, f(x,y), is
determined by two functions - The amount of light falling on a scene, i(x,y)
- The reflectivity of the various surfaces in the
scene, r(x,y) - These two functions combine to get f(x,y)
i(x,y) r(x,y)
55Image Capture
- The range is practically bounded by the hardware
resolution. It is calibrated to 0 for black and
to 255 for white. Intermediate values are
different intensity of grey.
56Image Capture
Red sensor
Green sensor
Blue sensor
A colour image is formed by combining the 3
images captured by the red, blue, and green
sensors.
57Cameras
- 3 main types of camera
- Vidicons, charge coupled devices (CCDs), and
Complementary Metal Oxide Silicon (CMOS). - Vidicons
- Image is formed on a screen.
- The screen is scanned across to produce
continuous voltage signal which is proportional
to the light intensity of image.
58Cameras
- CCD sensors
- A 2D array of pixel photosensitive sensors
- The number of charge sites gives the resolution
of the CCD sensor - Each sensor generates charges proportional to the
incident light intensity. - Each column of sensors is emptied into a vertical
transport register (VTR). - The VTRs are emptied pixel by pixel into a
horizontal transport register (HTR). - HTR empties the information row by row into a
signal unit
59Cameras
- CMOS sensors
- A 2D array of lattice
- Charge incident on a site is proportional to the
brightness at a point - Charge is then read like computer memory
- A particular site is selected by the control
circuit and the content of the site is read - The number of charge sites gives the resolution
of the CCD sensor
60Comparison of Camera Sensors
- Vidicon
- ? Cheap
- ? Aging problem due to moving parts
- ? Delay in response to moving objects in a scene
- CCD
- ? Irregularity in charge sitess material
(silicon) - ? Blooming effect because size is limited to 4 µm
- CMOS
- ? Irregular in charge sitess material (silicon)
- ? Directly related to intelligent cameras with
on-board processing - ? Cheaper than CCD
- ? Small size at around 0.1 µm
61Image Quality Measurements
- Image measurements processes include
- Imaging devices focus radiant energy onto a
photosensitive detector - The quantity of energy being absorbed is measured
digitally - The measurements are assembled into an image
- All these processes are subject to inaccuracies
gt images are not strictly accurate.
62Source of degradation
- Refraction of light at lens
- Distortion by camera lens
- Scattering of light by transmission medium in the
light path - Imperfect detector
- Blooming effect
63Refraction
- Lens are made of glass that refracts light
- The amount of refraction depends on the
wavelength of the light - A ray of white light is split into the
constituent colours.
64Refraction
- An optical system, based on lenses, cannot focus
light from a point onto a fixed point. - ? White objects have a coloured fringe.
- ? Sharp boundaries are blurred
65Geometric Distortion
- The best focus of scene is on a spherical surface
behind the lens - It is difficult, if not impossible, to fabricate
spherical detectors - Image projected onto a planar detector
- Severe distortion when focal length of lens is
short
66Geometric Distortion
- Image of rectangular grid captured using a short
focal length lens
67Scattering
- Light rays are dispersed by materials between the
reflected surface and the detector - E.g. fog disperse light
68Imperfect Detector
- Sensitivity of detector is not uniform
- Noise in the electronics that converts radiant
energy to an electrical signal - Usually quoted by camera manufacturers
- Overflow or clipping of signals when the signal
is of excessive amplitude. - The signal amplitude overflows the max value
- The signal amplitude might be clipped
- The detector is calibrated at each pixel
69Imperfect Detector
- Underexposure image lose image details
70Imperfect Detector
- Underexposure image lose image details
71Electronic Noise
- Noise is defined as deviation of the signal from
its expected value - Electronic noise produces speckle effect
- Two types of speckle effect
- Salt and pepper noise Random white and black
noise on image - Gaussian noise Random noise superimposed on the
signal being captured
72Salt and Pepper Noise
73Blooming effect
- A bright point source of light
- Cameras optical system will blur image slightly
- Neighbouring cells in the detector will respond
to this received impulse - ? Inaccurate brightness at neighbouring pixels
- ? when auto exposure is used, image details are
lost in other under-illuminated regions
74Blooming effect
Grey value
Pixel position
original grey value
Pixel position
75Image Quality Measurement
- Objective measurements
- Measured by instruments
- Invariant to the change of subjects
- Peak Signal to Noise Ratio (PSNR)
- Subjective measurements
- Measured by human beings
- Variant to the change of subjects
- Human survey
76Signal-to-Noise Ratio
- Signal-to-Noise Ratio (SNR) is the ratio of the
signal to the noise - It is often measured in decibels (dB) as
77Signal-to-Noise Ratio
- The Peak Signal-to-Noise Ratio (PSNR) is often
considered. Thus, the ratio of the maximum value
of the signal to the measured noise amplitude (in
dB).
78Image Resolution
- 3 resolutions
- Spatial resolution (no. of pixels)
- Brightness (no. of greylevels)
- Temporal (number of frames per second)
- These resolutions do have a mutual dependency
79Spatial Resolution
- An image is represented as a two-dimension array
of sample points called pixels. - E.g. A 320 x 200 image has 320 pixels on each
horizontal line and 200 pixels on each vertical
line.
320
x x x x x x x x x x x x x x x x x x x x x x x x x
x x x x x x x x x x x x x x x x x x x x x x x x x
x x x x x x x x x x x x x x x x x x x x x x x x x
x x x x x x x x x x x x x x x x x x x x x x x x x
x x x x x x x x x x x x x x x x x x x x x x x x x
x x x x x x x x
pixel
200
80Spatial Resolution
- A simple definition of spatial resolutions
- spatial resolution h v
- where h no. of horizontal pixels and v no.
of vertical pixels.
81Spatial Resolution
- Field of view, ?v the angle subtended by rays
of light that hit the detector at the edge of the
image - Relationship between field of view, the camera
detector, and focal length, f.
h
?v
v
f
82Spatial Resolution
- Horizontal Angular Resolution (HAR) is the ratio
of horizontal field of view, ?h, divided by the
number of pixels across the image - Vertical Angular Resolution (VAR) is the ratio of
vertical field of view, ?v, divided by the number
of pixels of the image vertically -
83Spatial Resolution
- Usually, HAR VAR by design
- r the smallest resolvable object
- Z the distance from the camera
- ? the angular resolution in radians
84Spatial Resolution
- To resolve an object 2mm in diameter at a range
of 1m, the minimum angular resolution, ?, needs
to be - ? 2mm/1m
- 0.002 radian
- 0.1146 degree.
85Spatial Resolution
- According to Nyquist theorem, at least two
samples per period are needed to represent a
periodic signal unambiguously, - Applying the Nyquist theorem to the spatial
dimension, two pixels must span the smallest
dimension of an object in order for it to be seen
in the image.
86Spatial Resolution
- Lena images with different sample points 256x256
and 512x512 pixels
87Spatial Resolution
- When a fixed size displaying window shows an
image of varying resolution - Low resolution image loses details
- High resolution image shows details
Low resolution
High resolution
88Brightness Resolution
- For monochrome image,
- Brightness resolution Number of grey levels
- Our eyes can differentiate around 40 shades of
grey only - Image capture devices are limited in
differentiating number of grey levels - Most monochrome images are captured using 8 bit
values. Range of grey levels is 0, 255.
89Brightness Resolution
- Images with more shades of grey show more details
bw
8 bit grey
90Colour Resolution
- For colour images, a display device may use fewer
colours - Colour resolution number of distinguishable
colours - Colour images are captured using three eight bit
values.
91Colour Resolution
- Images with more colours show image with high
fidelity
24 bit colour
8 bit colour
92Image Bits per Pixel
- 1 bit/pixel black white image, facsimile image
- 4 bits/pixel computer graphics
- 8 bits/pixel greyscale image
- 16 or 24 bits/pixel colour images
- colour representations RGB, HSV, YUV, YCbCr
93Interactions
- Spatial and brightness resolution can interact in
affecting the overall perception of an image - Poor spatial resolution may be compensated for by
a good brightness resolution - For each brightness resolution, a threshold
spatial resolution can be defined. Images with
higher spatial resolution are acceptable whereas
images with lower spatial resolution are
unacceptable.
94Interaction
8 bit greyscale
50x50
25x25
100x100
200x200
95Interaction
8 bit 256 colours
25x25
50x50
100x100
200x200
96Interaction
24 bit colour
25x25
50x50
100x100
200x200