W5 structuring two-dimensional-space
- 1. Visual Thinking and Information Visualization Design 10/18 Week5
- 3. • Patterns are formed where ideas in the head meet the evidence of the world • Top-down attentional processes cause patterns to be constructed From the retinal image that has been decomposed into a myriad of features of V1 Patterns are formed in a middle ground of fluid dynamic visual processing that pulls meaningful patterns Impose order Patterns are also the building blocks of objects p.44
- 4. 2.5D SPACE p.46
- 5. THE BINDING PROBLEM: FEATURES TO CONTOURS The process of combining different features that will come to be identified as parts of the same contour or region is called binding. Binding is essential because it is what makes disconnected pieces of information into connected pieces of information. p.46-47
- 6. THE BINDING PROBLEM: FEATURES TO CONTOURS How the responses of individual feature-detecting neurons may become bound into a pattern? Image we are a single neuron in the primary visual cortex that responds to oriented edge information, we get most excited when part of an edge or a contour falls over a particular part of the retina to which we are sensitive. p.47
- 7. THE BINDING PROBLEM: FEATURES TO CONTOURS To form part of our edge-detecting club, they must response to part of the visual world that is in-line with our preferred axis, and they must be attuned to roughly the same orientation. p.47-48
- 8. THE BINDING PROBLEM: FEATURES TO CONTOURS Our neuron is also surrounded by dozens of other neurons that respond to other orientations and it actively discourages them. Neurons responding to little fragments of edges have their signals suppressed. p.48
- 9. THE BINDING PROBLEM: FEATURES TO CONTOURS Binding is not only something required for the edge-finding machinery. A binding mechanism is needed to account for how information stored in various parts of the brain is momentarily associated through the process of visual thinking. p.49
- 10. THE GENERALIZED CONTOUR The brain requires a generalized contour extraction mechanism in the pattern-processing stage of perception. Emerging evidence points to this occurring in a region called the lateral occipital cortex (LOC) that receives input from V2 and V3 allowing us to discern an object or pattern from many kinds of visual discontinuity. p.49
- 11. TEXTURE REGIONS The edges of objects in the environment are not always defined by clear contours. The brain contains mechanisms that rapidly define regions having common texture and color. p.50
- 12. TEXTURE REGIONS The left- and right- hand sides of these texture patches are easy to discriminate because they contain differences in how they affect neurons in the primary visual cortex. p.50
- 13. TEXTURE REGIONS When low-level feature differences are not present in adjacent regions of texture, they are more difficult to distinguish. The following examples show hard-to-discriminate texture pairs. p.51
- 14. TEXTURE REGIONS Mutual excitation between similar features in V1 and V2 explains why we can easily see regions with similar low-level feature components through spreading activation. Only basic texture differences should be used to divide space. p.51
- 15. INTERFERENCE AND SELECTIVE TUNING To minimize this kind of visual interference, one must maximize feature-level differences between patterns of information. Having one set of features move, and another set static is probably the most effective way of separating overlaying patterns. p.51
- 16. PATTERNS, CHANNELS, AND ATTENETION Some pattern-level factors, such as the smooth continuity of contours are also important. Once we choose to attend to a particular feature type, such as the thin red line in illustration below, the mechanisms that bind the contour automatically kick in. p.52
- 17. PATTERNS, CHANNELS, AND ATTENETION If different zones can be represented in ways that are as distinct as possible in terms of simple features, the result will be easy to interpret, although it may look stylistically muddled. p.52-53
- 18. INTERDIATE PATTERNS Each of the regions provides a kind of map of visual space that is distorted to favor central vision. As we move up the hierarchy the mapping of visual imagery on the retina becomes less precise. p.53
- 19. INTERDIATE PATTERNS Currently, there is no method for directly finding out which of an infinite number of possible patterns a neuron responds to best. As a result much less is known about the mid-complexity patterns of V4 than the simple ones of V1 and V2. What is known is they include patterns such as spikes, convexities, concavities, as well as boundaries between texture regions, and T and X junctions. p.53
- 20. INTERDIATE PATTERNS In addition to static patterns, V4 neurons also respond well to different kinds of coherent motion patterns. There is emerging evidence that biomotion perception may have its own specialized pattern detection mechanisms in the brain. p.54
- 21. PATTERN LEARNING As we move up to the processing chain from V1 to V4 and on to the IT cortex, the effects of individual experience become more apparent. All human environments have objects with well-defined edges, and this common experience means that early in life everyone develops a set of simple oriented feature detectors in their V1. At the level of V1 everyone’s pattern is pretty much the same, and this means that design rules based on V1 properties will be universal. There is a critical period in the first few years of life when the neural pattern-finding systems develop. • The order we get, the worse we are at learning new patterns, and this seems to be especially true for the early stages of neural processing. p.54
- 22. SERIAL PROCESSING The more elaborate patterns we encounter at higher levels in the visual processing chain are to a much greater extent the product of an individual’s experience. All visual thinking is skilled and depends on pattern learning. The map reader, the art critic, the geologist, the theatre lighting director, the restaurant chef, the truck driver, and the meat inspector have all developed particular pattern perception capabilities encoded especially in the V4 and IT cortical areas of their brains. For the beginning reader, a single fixation and a tenth of a second may be needed to process each letter shape. The expert reader will be able to capture whole words as single perceptual chunks if they are common. This gain in skill comes from developing connections in V4 neurons that transforms the neurons into efficient processors of commonly seen patterns. p.55
- 23. VISUAL PATTERN QUERIES AND THE APPREHENDABLE CHUNK Patterns range from novel to those that are so well learned that they are encoded in the automatic recognition machinery of the visual system. They also range from the simple to the complex. Apprehendability is not a matter of size so long as a pattern can be clearly seen. For unlearned patterns the size of the apprehendable chunk appears to be about three feature components. p.55
- 24. VISUAL PATTERN QUERIES AND THE APPREHENDABLE CHUNK For unlearned patterns the size of the apprehendable chunk appears to be about three feature components. p.55
- 25. MULTI-CHUNK QUERIES Performing visual queries on patterns that are more complex than a single apprehendable chunk requires substantially greater attentional resources and a series of fixations. When the pattern is more complex than a single chunk the visual query must be broken up into a series of subqueries, each of which is satisfied, or not, by a separate fixation. p.56
- 26. SPATIAL LAYOUT Pattern perception is more than contours and regions. Group of objects can form patterns based on the proximity of the elements. p.56
- 27. SPATIAL LAYOUT We do not need to propose any special mechanism to explain these kinds of grouping phenomena. A group smaller points or objects will provoke a response from a large-shape detecting neuron, while at the same time other neurons, tuned to detect small features, respond to the individual components. p.56
- 28. SPATIAL LAYOUT Objects in the real world have structure at many scales. p.57
- 29. SPATIAL LAYOUT V4 neurons still respond strongly to patterns despite distortions that stretch or rotate the pattern, so long as the distortion is not too extreme. A single V4 neuron might respond well to all of these different T patterns. p.57
- 30. HORIZONTAL AND VERTICAL We are very sensitive to whether something is exactly vertical or horizontal and can make this judgment far more accurately than judging if a line is oriented at forty-five degree. The reason for this has to do with two things: 1. One is to maintain our posture with respect to gravity; 2. The other is that our modern world contains many vertically oriented rectangles and so more of our pattern-sensitive neurons become tuned to vertical and horizontal contours. This make the vertical and horizontal organization of information an effective way of associating a set of visual objects. p.57
- 31. PATTERN FOR DESIGN Most designs can be considered hybrids of learned symbolic meanings, images that we understand from our experience, and pattern that visually establish relationships between the components, trying the design elements together. A design can be made visually efficient by expressing relationships by means of easily apprehended patterns. p.58
- 32. PATTERN FOR DESIGN Relationships between meaningful graphical entities can be established by any of the basic pattern-defining mechanisms. p.58
- 33. PATTERN FOR DESIGN Graphic patterns defined by contour or region can be used in very abstract ways to express the structure of ideas. p.58
- 34. PATTERN FOR DESIGN Relationship defining patterns can also be used in combination. The design challenge is to use each kind of design device to its greatest advantage in providing efficient access to visual queries. p.59
- 35. PATTERN FOR DESIGN Relationship defining patterns can also be used in combination. The design challenge is to use each kind of design device to its greatest advantage in providing efficient access to visual queries. p.59
- 36. EXAMPLES OF PATTERN QUERIES WITH COMMON GRAPHICAL ARTIFACTS The point of the following examples has been to show that a wide variety of graphical display can be analyzed in terms of visual queries, each of which is essentially a visual search for a particular set of patterns. p.60
- 37. p.60
- 38. p.61
- 39. EXAMPLES OF PATTERN QUERIES WITH COMMON GRAPHICAL ARTIFACTS p.62
- 40. EXAMPLES OF PATTERN QUERIES WITH COMMON GRAPHICAL ARTIFACTS p.62
- 41. SEMANTIC PATTERN MAPPING For the most part, when we see patterns in graphic designs we are relying on the same neural machinery that is used to interpret the everyday environment. There is, however, a layer of meaning—a kind of natural semantics—that is built on top of this. For example, we use a big graphical shape to represent a large quantity in a bar chart. We use something graphically attached to another object to show that it is part of it. These natural semantics permeate our spoken language, as well as the language of design. p.62
- 42. SEMANTIC PATTERN MAPPING Here we are concerned with spatial metaphors and their use in graphic design, where they are just as ubiquitous. p.63
- 43. SEMANTIC PATTERN MAPPING In this chapter, we have used the term pattern in its everyday sense to refer to an abstract arrangement of features or shapes. There is a more general sense of the word used in cognitive neuroscience, according to which every piece of stored information can be thought of as a pattern. Thus complex patterns are patterns of patterns. Finding patterns is the essence of how we make sense of the world. In fundamental sense, the brain can be thought of as set of interconnected pattern-processing modules. The processing of visual thinking also involves the interplay between patterns of activity in the non-visual verbal processing centers of the brain. p.63
- 44. SEMANTIC PATTERN MAPPING p.63-64