vision - unimi.it
TRANSCRIPT
Vision
Stefano Ferrari
Universita degli Studi di [email protected]
Methods for Image processing
academic year 2015–2016
Biological vision and image processing
I The human visual perception results from a chain ofprocessing tasks, often cooperating with other perceptualactivities.
I Image processing is a discipline (mainly) based onmathematical and statistical concepts, with constraintsimposed by the image acquisition technology.
I The study of the human vision is useful for more than onereason:
I the output of some processes is targeted for the vision of ahuman user;
I the operation of the organs devoted to the acquisition andperception of images can inspire the development of similarartificial devices; moreover, the performance of imageprocessing algorithms have to be compared;
I some known mulfunctioning of the biological apparatus do notaffect the image processing devices.
.
Stefano Ferrari— Methods for Image processing— a.a. 2015/16 1
Eye
I cornea
I sclera
I pupil
I iris
I lens
I choroid
I retina
I fovea
I blind spot
Blind spot: close the left eye, sta-re to the cross, come closer to thescreen till the spot disappears.
Distribution of the receptors on the retina
On the retina, twokinds of receptorscan be found:
I cones;
I rods.
I Cones:I about 7 millions;I perception of details and rapid changes;I sensitive to colors;I sensitive in high illuminance conditions (photopic vision).
I Rods:I about 100 millions;I provide a large scale vision;I sensitive in low illuminance conditions (scotopic vision).
.
Stefano Ferrari— Methods for Image processing— a.a. 2015/16 2
Image formation
I The distance the lens and the retina is fixed (about 17 mm).
I The scene is focused by changing shape of the lens (fordistances up to three meters).
I The dimension of the object onto the retina is a function ofits distance (no zoom!).
Perceived brightnessI The human eye can operate in
a very large range of lightintensity level, from scotopicvision to glare.
I 1010 discriminated levels.I Perceived brightness is a
logartimic function of thelight intensity.
I However, this interval is notcovered simultaneously, butonly a subset of levels close tothe brightness of the scene canbe discriminated.
I The sensitivity range is shiftedas the brightness changes(brightness adaptation).
.
Stefano Ferrari— Methods for Image processing— a.a. 2015/16 3
Perceived brightness (2)
I The perceivedbrightness (subjectivebrightness) dependson the environmentalbrightness, Ba.
I Under the lightintensity thresholdBb, black isperceived.
I The upper limit is lessmeaningful: abrighter light canshift Ba.
Discrimination of intensities
I Experiment for evaluating thesensitivity to brightness changes: on auniformly illuminated screen atintensity I , a short-duration spot isprojected at intensity I + ∆I .
I As ∆I is progressively increased, thespot is more esily perceived: ∆Ic is theintensity change that makes the spotdiscriminated at 50% of the time.
I ∆Ic/I is called Weber ratio.
I A high value means low sensitivity: thespot have to be much brighter thanthe background to be discriminated.
.
Stefano Ferrari— Methods for Image processing— a.a. 2015/16 4
Discrimination of intensities (2)
a
b
I The eye can discriminate few(10–25) shades with respect tothe environmental brightness.
I However, this does not meansthat all the images can bereported using few shades:maybe locally they can beenough, but the brightness ofthe scene can change.
The same image at 256 (a) and 24(b) gray levels. In the regions whe-re the intensity changes slowly (skyand clouds), the lack of intermediateshades is apparent.
Perceived intensity
I The perceived brightnessdoes not depends only bythe intensity.
I The Mach bands areconstant intensity region.
I However, the boundarybetween two bands seemsdarker in the darkerregion and brighter in thebrighter band side.
.
Stefano Ferrari— Methods for Image processing— a.a. 2015/16 5
Contrast
a
b
I The phenomenon calledsimultaneous contrast isanother clue that theperceived brightness is notonly a function of theintensity.
I The center squares of figure(a) have the same intensity,but the perceived brightnessdepends on the backgroundintensity.
I Similarly, despite it seemsimpossible, in figure (b) thesquares A and B share thesame intensity.
Optical illusions
a b
c d
I The optical illusions areperceptions that mistakenlyrepresent the observedreality.
I The are caused bypeculiarities of the humanvision system and are aphenomenon activelystudied.
(a) e (b) perception of inexistentshapes. (c) wrong estimation ofthe length. (d) diagonal parallellines perceived as non-parallel.
.
Stefano Ferrari— Methods for Image processing— a.a. 2015/16 6
Optical illusions (2)
The optical illusions can occur also at higher level of abstraction.
The ability of recognizing shapes under vague stimulation (e.g.,animal-like shaped clouds) is well known and is called pareidolia.
Optical illusions (3)
a b
c d
The human visualsystem operates fa-ce recognition veryefficiently.The perceived dif-ference between theimages (a) and (b)is less than thatbetween (c) and (d).
.
Stefano Ferrari— Methods for Image processing— a.a. 2015/16 7
Optical illusions (4)
The perceived expression changes with the distance.
We see with the brain
I The optical illusions are very variegated and include bothstatic and dynamic phenomena.
I A good collection (with explanations):I 84 Optical Illusions & Visual Phenomena by Michael Bach,
http://www.michaelbach.de/ot.
I The wrong perception phenomena also affects the othersenses. For instance, the third sound of Tartini and binauralsounds are well known audio illusions.
.
Stefano Ferrari— Methods for Image processing— a.a. 2015/16 8
Homeworks and suggested readings
DIP, Section 2.1
I pp. 35–43
Mirage with parabolic mirrors
I http://www.ap.smu.ca/demos/index.
php?option=com_content&view=
article&id=122&Itemid=85
Autostereograms
I http://en.wikipedia.org/wiki/
Autostereogram
.
Stefano Ferrari— Methods for Image processing— a.a. 2015/16 9