Perceiving Recognizing Objects 1 Chapter 4 1 What is middle vision a Middle Vision Pg 97 loosely defined stage of visual processing that comes after basic features have been extracted from the image low level early vision up to striate and before object recognition and scene understanding high level vision i Goal is to organize elements of a visual scene into groups that we can then recognize as objects ii Like a collection of specialists perceptual committees that each have an area of expertise and individual opinions about what an input means the goal is to get one answer from lots of opinions b Finding edges i We cant just group all edges that touch one another into one object because objects overlap and poke out from other objects but with some shapes you can look at them with missing pieces in their edges and still know what they are 1 In the early stages of processing the visual system can figure out which edges mark the boundary of an object and which represent surface features these pieces of information are combined together to make the systems best guess about the presence of contour a Illusory contours Pg 98 contour that s perceived even though nothing changes from one side of it to the other in an image 2 Why do you think the Gestalt school of thought was instrumental in identifying how we perceive images To answer this you need to know what Gestalt means a Next Question 3 Know the gestalt grouping principles good continuation occlusion texture segmentation similarity proximity connectedness common region how do those latter 2 relate to proximity what trumps what parallelism and symmetry Also know dynamic occlusion a Gestalt Pg 99 school of thought stressing that the perceptual whole could be greater than the apparent sum of its parts the whole is greater than the sum of its parts an image will appear to group together set of rules describing which elements in b Gestalt Grouping Rules Pg 99 i These rules are useful because they reflect the regularities of the world they allow the visual system to say if the input looks like this I can infer that this is the state of the world 1 More simply you can take inputs that are regular of the ii Good Continuation Pg 100 objects seen in the world and you can make really educated inferences on what they are based on these rules gestalt grouping rule stating that two elements will tend to group together if they seem to lie on the same contour 1 When you see an X that doesn t have completed lines you ll be able to figure out which lines continue in what direction a Two pieces of an edge make it easier to perceive a third line in between them even if there isn t one lines of similar orientation go together If the set of lines form a closed shape like a square it ll be even easier to assume the missing pieces b i Closure Pg 100 Gestalt principle that holds that a closed contour is preferred to an open contour 2 Color trumps good continuations as an organizing principle a You ll assume that line pieces of the same color go together even if they appear to be more inline with a piece of a line that s a different color iii Occlusion If an edge seem to stop suddenly the visual system guesses that its stopping because something else is in the way occluding it 1 Dynamic Occlusion when visual regions are discontinuous not only in space but over time iv Texture Segmentation Pg 101 common texture properties carving an image into regions of 1 Connecting line segments aren t enough for us to divide a raw image into recognizable objects a striate edge detector wont be able to tell the difference between textures because they only focus on edges 2 Can be based on similar color size orientation and aspects of form symmetry 3 Closely related to similarity proximity parallelism and v Similarity Pg 102 gestalt grouping rule stating that the gestalt grouping rule stating that the tendency of two features to group together will increase as the similarity increases vi Proximity Pg 102 tendency of two features to group together will increase as the distance between them decreases be grouped together common region they will be grouped together if two features appear to be part of the same if two features appear to be connected they will vii Connectedness viii Common Region 1 Common region and connectedness trump proximity 4 How does color fit in to the gestalt grouping principles a The previous question 5 What is an ambiguous figure a Ambiguous figure Pg 104 interpretations of its identity or structure generates two or more plausible interpretations visual stimulus that gives rise to two or more i Every image is theoretically ambiguous but the perceptual committees almost always agree on a single interpretation 1 In this case two committee members are arguing leading us to struggle with recognition see slide 19 for an image as an example 6 What is an accidental viewpoint a Accidental viewpoints Pg 105 regularity in the visual image that isn t present in the real world viewing position that produces some i Like two sides of independent objects lining up perfectly usually the visual system doesn t allow this its too smart so it just doesn t consider the option 7 What is figure and ground and what principles do we use to decide which is which in a visual scene What is evidence that maybe object recognition precedes establishment of figure ground a Figure ground assignment Pg 106 regions of an image belong to a foreground object figure and others are part of the background ground process of determining that some b Principles i Surroundedness Pg 106 ii Size iii Symmetry iv Parallelism the smaller region is likely to be the figure rule for figure ground assignment stating that if one region is entirely surrounded by another its likely that the surrounded region is the figure the figure determine which portion of a display is the foreground and which portion is the background symmetrical regions are more likely to be the figure regions with parallel contours are more likely to be how surface details more relative an edge can v Relative motions 1 If certain parts of an object or scene move together it tells us they go together which tells us they re part of the foreground vi Synchrony vii External edges elements that change at the same time group together tangent to the surface the shading of objects tells us about figure vs ground the set of surface points whose sight lines are 1 Something that s larger making it look closer has 3D shading or just is 3D will be
View Full Document
Unlocking...