ee 576 - introductionisl.ee.boun.edu.tr/courses/ee576/lectures/sunum/intro.pdf · 2019. 2. 13. ·...
TRANSCRIPT
Outline
Introduction
EE 576 - Introduction
H.I. Bozma
Electric Electronic Engineering
Bogazici University
February 13, 2019
H.I. Bozma EE 576 - Introduction
Outline
Introduction
IntroductionViewing: Biological VisionLightVision SystemsImage Transformations
H.I. Bozma EE 576 - Introduction
Outline
Introduction
Viewing: Biological Vision
Light
Vision Systems
Image Transformations
Vision
◮ Vision : Allow us or a machine to see.
◮ Generating a symbolic description from one or more image.
◮ How the brain processes visual information Relatively littleknown.
◮ Study computer-vision paradigm
◮ Highlight methods of computation that are in generalbiologically inspired.
◮ Engineering approach - a solution regardless of biologicalsystem.
H.I. Bozma EE 576 - Introduction
Outline
Introduction
Viewing: Biological Vision
Light
Vision Systems
Image Transformations
Factors Influencing Vision
H.I. Bozma EE 576 - Introduction
Outline
Introduction
Viewing: Biological Vision
Light
Vision Systems
Image Transformations
Factors Influencing Vision
◮ Viewer: Imaging sensor, properties and geometry
◮ Light:◮ Wavelength and color◮ Intensity◮ Contrast (I (x)− I (y))/(I (x) + I (y))◮ Polarization – Not part of human vision, but commonplace in
animal world
◮ Physical properties including geometry, shape and material
◮ Visual processing
H.I. Bozma EE 576 - Introduction
Outline
Introduction
Viewing: Biological Vision
Light
Vision Systems
Image Transformations
Eye
◮ Organ providing spatial vision.
◮ The information collected → Processed by the brain.
◮ The brain - Visualize the world as consisting of gratings of acomplete range of spatial frequencies.
H.I. Bozma EE 576 - Introduction
Outline
Introduction
Viewing: Biological Vision
Light
Vision Systems
Image Transformations
Human Vision Components
H.I. Bozma EE 576 - Introduction
Outline
Introduction
Viewing: Biological Vision
Light
Vision Systems
Image Transformations
Evolution of the Eye
◮ Difficult and time consuming
◮ Gradual development from the simplest form of sensitivity tofocused eye =⇒ 530 million years
◮ Driving force =⇒ Better resolution
◮ The development of eyes =⇒ Increases in size, speed andarmour
◮ Did eyes evolve once or many times?
H.I. Bozma EE 576 - Introduction
Outline
Introduction
Viewing: Biological Vision
Light
Vision Systems
Image Transformations
Good Eye?
◮ Retinal sampling frequency
◮ Optical cutoff – As details become finer, difficult to resolve.For example, leaves of a distant tree lose their identity in theoverall texture.
H.I. Bozma EE 576 - Introduction
Outline
Introduction
Viewing: Biological Vision
Light
Vision Systems
Image Transformations
Eye Resolution
H.I. Bozma EE 576 - Introduction
Outline
Introduction
Viewing: Biological Vision
Light
Vision Systems
Image Transformations
Image Formation
H.I. Bozma EE 576 - Introduction
Outline
Introduction
Viewing: Biological Vision
Light
Vision Systems
Image Transformations
Comparison of Resolutions
H.I. Bozma EE 576 - Introduction
Outline
Introduction
Viewing: Biological Vision
Light
Vision Systems
Image Transformations
Good Eye?
◮ Retinal sampling frequency
◮ Optical cutoff – As details become finer, difficult to resolve.For example, leaves of a distant tree lose their identity in theoverall texture.
H.I. Bozma EE 576 - Introduction
Outline
Introduction
Viewing: Biological Vision
Light
Vision Systems
Image Transformations
Human Eye Stabilization
H.I. Bozma EE 576 - Introduction
Outline
Introduction
Viewing: Biological Vision
Light
Vision Systems
Image Transformations
Vestibulo-ocular Reflex (VOR)
◮ A reflex eye movement that stabilizes images on the retinaduring head movement by producing an eye movement in thedirection opposite to head movement, thus preserving theimage on the center of the visual field.
◮ When the head moves to the right, the eyes move to the left,and vice versa.
◮ Since slight head movement is present all the time, the VORis very important for stabilizing vision:
◮ No dependency on visual input and works even in totaldarkness or when the eyes are closed.
H.I. Bozma EE 576 - Introduction
Outline
Introduction
Viewing: Biological Vision
Light
Vision Systems
Image Transformations
OptoKinetic Reflex (OKR)
◮ Allows the eye to follow objects in motion when the headremains stationary
H.I. Bozma EE 576 - Introduction
Outline
Introduction
Viewing: Biological Vision
Light
Vision Systems
Image Transformations
Optical Defects
H.I. Bozma EE 576 - Introduction
Outline
Introduction
Viewing: Biological Vision
Light
Vision Systems
Image Transformations
Environmental Light and Sensitivity
Figure: Environmental light and sensitivity (a) Sun (b) Reflectance offour flowers, (c) Sensitivity of human rods and cones, (d) Sensitivity ofbee photoreceptors .
H.I. Bozma EE 576 - Introduction
Outline
Introduction
Viewing: Biological Vision
Light
Vision Systems
Image Transformations
Color and Wavelength
◮ Human visible range 400− 700nm and sensitivity to 800nm
◮ Animals – Extend into ultraviolet range 400− 320nm
◮ Animals – Infrared radiation given off by hot bodies – Thermalimaging (Many animals such as snakes, beetles)
◮ Categorization of objects wrt to reflected light frequency◮ Leaves – 500-600nm, Blue flowers – 350-500nm, Ripe fruit –
550-600nm
◮ Differentiate btw wavelength and color◮ Need at least visual pigments or some form equivalent◮ Typically three visual pigments◮ Some even more – 12 in one case.◮ Learn not just the peak sensitivities, but also the intermediate
wavelengths and wavelength combinations (using stimulationratios)
H.I. Bozma EE 576 - Introduction
Outline
Introduction
Viewing: Biological Vision
Light
Vision Systems
Image Transformations
Visual Pigments
Figure: Two visual pigments.H.I. Bozma EE 576 - Introduction
Outline
Introduction
Viewing: Biological Vision
Light
Vision Systems
Image Transformations
Eye - How reliable is derived intensity?
Figure: Are the two intensities the same?
H.I. Bozma EE 576 - Introduction
Outline
Introduction
Viewing: Biological Vision
Light
Vision Systems
Image Transformations
Eye - How reliable is derived geometry?
Figure: Geometry?
H.I. Bozma EE 576 - Introduction
Outline
Introduction
Viewing: Biological Vision
Light
Vision Systems
Image Transformations
Vision Systems
Regardless of natural or artificial Same basic components:
◮ Imaging devices + interface systems
◮ Visual Processing – Information about surroundings thatenables us to interact intelligently with the environment.
H.I. Bozma EE 576 - Introduction
Outline
Introduction
Viewing: Biological Vision
Light
Vision Systems
Image Transformations
Comparison
H.I. Bozma EE 576 - Introduction
Outline
Introduction
Viewing: Biological Vision
Light
Vision Systems
Image Transformations
Really difficult problems for machine vision?
Figure: Are the two intensities the same?
H.I. Bozma EE 576 - Introduction
Outline
Introduction
Viewing: Biological Vision
Light
Vision Systems
Image Transformations
Imaging Devices
◮ Photographic imaging
◮ Scanners
◮ Cameras
◮ Laser range finders
◮ Ultrasound
◮ Reconstruction imaging (PET, MRI)
◮ RGB-D devices (Kinect, etc.)
H.I. Bozma EE 576 - Introduction
Outline
Introduction
Viewing: Biological Vision
Light
Vision Systems
Image Transformations
Image Examples
H.I. Bozma EE 576 - Introduction
Outline
Introduction
Viewing: Biological Vision
Light
Vision Systems
Image Transformations
Image Acquisition & Processing Hardware
◮ Imaging◮ Resolution : Spatial resolution (512 × 480, 640 × 480)◮ Aspect ratio: The ratio of horizontal vs vertical elements◮ Memory:
◮ Store & display images
◮ Processing: Hardware and software
H.I. Bozma EE 576 - Introduction
Outline
Introduction
Viewing: Biological Vision
Light
Vision Systems
Image Transformations
Image Representation
◮ Imaging devices Incoming scenery → Single image ortime-varying image I .
◮ Image plane X ⊂ R × R as I : X → C where C is the colorspace.
◮ x ∈ X → I (x1, x2) ∈ C .
◮ Time-varying image → I : X × R → C as I (x1, x2, t) ∈ C .
◮ C represented digitally – n − bit resolution.◮ In most cases, n = 8 – 28 = 256 color levels.◮ Signal (light, ultrasound or laser)◮ In case of light - whether gray or color images are considered.◮ In the case of gray images, C = {0, · · · , 255}.◮ In case of color images, the color space has three components
– corresponding to red, green and blue components asC = {0, · · · , 255} × {0, · · · , 255} × {0, · · · , 255}.
H.I. Bozma EE 576 - Introduction
Outline
Introduction
Viewing: Biological Vision
Light
Vision Systems
Image Transformations
Vision Paradigm
Figure: Classical vision paradigmH.I. Bozma EE 576 - Introduction