Visual Perception PDF Lecture Notes

Visual Perception KMF 1023 COGNITIVE PSYCHOLOGY Lecturer: Norehan Zulkiply FSKPM Universiti Malaysia Sarawak In this topic, we will learn about: Perception...

Visual Perception KMF 1023 COGNITIVE PSYCHOLOGY Lecturer: Norehan Zulkiply FSKPM Universiti Malaysia Sarawak In this topic, we will learn about: Perception Perception – How does the process of perception take place? The conscious experience that – Two important processes: bottom-up vs top-down process results from stimulation of the Bottom-up processing: – How do we analyse objects into features in the early process senses of perception? FIT Theory vs RBC Theory Top-down Processing: – How are elements of a scene organised into objects? Gestalt Approach Evidence that perception is an intelligent activity – Some behavioral and physiological evidence 1  How does Perception take place? How does Perception take place? Why is something easy like looking at a Two important basic processes in scene becomes complicated when we look perception at the mechanisms involve? – Bottom-up processing – Top-down processing How does our brain store information about our perception of the world? How does Perception take place? How does Perception take place? Bottom up processing Top-down processing – Data based processing – Knowledge-based processing – Processing based on data received from – Processing that is based on knowledge stimulus (eg. light) – Knowledge of a context can influence one’s – No incoming data, no perception perception The case of Misperception The case of Misperception (Interpretation) (Transduction & Transformation) Light from the streetlight is reflected from the sign into Roger’s eye, which The pattern on the retina is transformed into electrical signals, which are creates a pattern representing the street sign on Roger’s retina. transmitted to the brain and processed. 2  Data + Knowledge = Perception Demonstration of bottom-up and top-down processing The rat-man demonstration Divide class into 2 groups – One group to see stimulus A, another group sees stimulus B – Both groups are then shown Picture A Roger’s perception of the sign is created by the processing of the incoming data provided by the pattern of light entering his eye (bottom-up processing) plus the influence of existing knowledge and expectations (top-down processing). Demonstration of bottom-up and Demonstration of bottom-up and top-down processing top-down processing What do you see in this picture ? Stimulus A Demonstration of bottom-up and Demonstration of bottom-up and top-down processing top-down processing What do you see in this picture ? Stimulus B Why do the two different groups give different answers? 3  Demonstration of bottom-up and Demonstration of bottom-up and top-down processing top-down processing Experiment by Stephen Palmer (1975): Effect of context on Stimulus A perception – data – presented a context scene such as on the left & then briefly – starting point for bottom up processing flashed one of the target pictures on the right When you look at the picture after you have looked at Stimulus A, an expectation is created – This expectation = knowledge = top-down processing – Top-down processing influences your perception of Picture A – “We see only what we know” – Speeded object-recognition task: – e.g. kitchen scene, then shown bread, mailbox, drum Demonstration of bottom-up and top-down processing Palmer’s experiment – Context-congruent objects: 80% recognition – Context-incongruent object: 40% Bottom Up Processing recognition - Shows how a person’s knowledge of a particular context provided by a scene can influence (or inhibit!) perception Bottom-up processing: Feature Bottom-up processing approach to object perception The first step in perceptual process Basic idea – We break down and analyse objects according to features before we put them together and synthesise them to recognise what objects Analysing objects into smaller components they are (features) – Analysis before synthesis – Feature approach to object perception Two theories E.g.: – Feature Integration Theory (FIT) A model for recognising letters/ alphabets – Recognition by Components approach (RBC) Example, you are shown the letter ‘A’ 4  Bottom-up processing: A model Bottom-up processing: A model for recognising letters for recognising letters To show how the analysis of features can lead to the recognition 2 stages of letters. Feature Analysis Stage Letter Analysis Stage Our mental bank consist of Our mental bank consist of letter feature units units Each unit respond to Each letter unit receives input from specific features the feature units associated with that letter ‘A’ activates three units: ‘A’ unit receives input from three feature units Other letters that have common A model for recognizing letters by analyzing their features. The stimulus, A, features with A also receive input activates three feature-units. These feature-units cause strong activation of the A letter-unit and weaker activation of units for letters such as the N and the T, from feature units that are activated which lack some of A’s features. The A is identified by the high level of activation by ‘A’ (eg. ‘A’, ‘N’, ‘T’) BUT only ‘A’ is of the A letter-unit. recognised due to a higher level of activation Bottom-up processing: A model for recognising letters Evidence for Feature Analysis Basic idea in feature analysis: Ann Treisman (1986)’s Visual Search Experiment – Activation of letter units provide the information to determine – Wants to find out how fast can people find a target which letter is present based on how many distractors are present? – Our visual system just needs to determine which unit is activated most strongly – Eg. Find the ‘O’ among the ‘V’s – Advantage of feature analysis detect shared features even when letters look different Easy to find the ‘O’ – Eg. as it “pops out” Even when there are many distractors – But the model would have difficulty with: More Evidence for Feature More Evidence for Feature Analysis Analysis Standard result for Visual Search experiment – The graph shows the time taken to identify the – Eg. Find the ‘R’ among the ‘P’s and ‘Q’s target & the no. of distractors exist. Difficult to find the ‘R’ Differences in results because Does not “pop out” of features of Takes a longer time the target letter and distractor when there are letter more distractors Shared features prevent “pop up” effect.  What are the shared features? Results of visual search experiments like the one in the demonstration. (a) Target = 0; Distractors = V; (b) Target = R; Distractors = P and Q. (Based on Treisman, 1986) 5  Permainan: Cuba cari Waldo 10/15/2021 10/15/2021 1) Feature Integration Theory (FIT) Evidence for FIT: Illusory Conjuctions Proposed by Treisman (1986) “Illusory” - misleading Treisman & Schmidt (1982)- showed evidence that objects are analysed in to features in the early stage of perceptual process. => Free floating features => Object Flow diagram for Treisman’s (1986) feature integration theory. According to this theory, objects are first analyzed into features in the preattentive stage, and then these features are combined into an object that can be Figure 3.18 (p. 68) perceived in the focused attention stage. Stimuli for Treisman and Schmidt’s (1982) illusory conjunction experiment. The geometrical figures were different colors, as indicated by the key. The numbers were black. Evidence for FIT: Illusory Conjuctions Evidence for FIT: Illusory Conjuctions Participants in the experiment shown Figure 3.18 for 20miliseconds Participants reported seeing “small red circle”, Features like ‘color’, ‘curvature’, ‘angle’ are free “small green triangle” floating Illusory Conjuctions – incorrect combination of features from two different stimulus Reason : – At the Preattentive stage-Components of perception exist independently of one another The results of the illusory conjunction experiment suggest that very early in the perceptual process, features that make up an object are “free floating.” This is symbolized here by showing some of the features of a cell phone as existing separately from one another at the beginning of the perceptual process. 6  Evidence for FIT: Illusory Conjuctions 2) Recognition By Components (RBC) Role of attention: Focus on shapes can eliminate illusory conjunctions Biederman (1987): – E.g. Participants were told to only focus on the objects and ignore the numbers – so all shapes were paired with their The basic component of correct colors perception are 36 geons Role of knowledge: Top-down processing can channel (3D shapes/volumes) perception – E.g Asked Ps to identify (colored, shaped) objects, the usual IC occurred. – But if participants were told that they will be looking at carrot, a lake and a tire, they were more likely to correctly identify the object and its colour Three stimuli used by Treisman to illustrate how top-down processing can Geons – features are not lines, curves or colors. influence the combining of features - e.g cylinders, pyramid, rectangular solids shapes 2) Recognition By Components (RBC) 2) Recognition By Components (RBC) Three properties of geons: Discriminability – View invariances – Each geon can be – Discriminability differentiated from others – Resistance to visual noise from all viewpoints (e.g cylinder vs. rectangular) View invariances (can be identified when view from diff. angles ) Resistance to visual noise – Occurs because geons have view invariant properties – Each geon can be identified even in noisy conditions – But when geons cannot be identified, we cannot recognise the object (e.g pg 72 - flashlight) 2) Recognition By Components (RBC) FIT vs RBC Conclusion from the RBC theory: Both agree on the idea of early analysis of objects into parts (in the early part of the If enough information is available for us to perception process) identify an object’s basic geons, we will be able to identify the object FIT – focus on basic features (lines, curves, colours) & the role of attention in We can recognise objects based on a small combining them number of basic shapes (e.g., when object view RBC – focus on how we perceive 3 is incomplete) dimensional shapes (geons) 7  Top-down Processing What do you see in the picture? Top Down Processing Top-down Processing The Gestalt Approach How is the environment organised into objects – The concept of Perceptual organisation - the Perception is due to laws of perceptual organisation of elements of the environments into objects organisation – 6 laws 1. Law of Pragnanz (good form) Perceptual organisation - Studied by Gestalt Psychologists 2. Law of Similarity Gestalt- German word for “form” The essential point in gestalt is that in perception, the 3. Law of Good Continuation whole is different from the sum of its parts. 4. Law of Proximity (or nearness) 5. Law of Common fate 6. Law of Familiarity During perception, the mind groups patterns according to Additional Laws: rules they called the laws of perceptual organization 7. Law of Closure 8. Law of Figure/Ground Gestalt Principles: Laws of Perceptual Gestalt Principles: Laws of Perceptual Organisation Organisation 1. Pragnanz / simplicity / good figure - A stimulus 2. Similarity: similar things are grouped together will be organized into as good a figure as possible – Elements that look similar will be perceived as Good = symmetrical, simple, and regular part of the same form. Example 1: Example 1: The above figure appears to the eye as a square overlapping triangle, not a combination of several shapes There seems to be a triangle in the square. Example 2: Example 2: 8 Gestalt Principles: Laws of Perceptual Laws of Perceptual Organisation Organisation 3. Good continuation – connected points form a straight or smooth Example 3: Grouping (due to similarity of curving line lightness) Example 1: People tend to draw a good continuous line. Points which, when connected, result in straight or smoothly curving lines, are seen as belonging together, and the lined tend to be seen as following the smoothest. Laws of Perceptual Organisation Laws of Perceptual Organisation 4. Proximity or nearness -Good continuation – things that are closer together will be perceived as Example 2: grouped together On the left, there appears to be three horizontal rows, while on the right, the grouping appears to be columns. Laws of Perceptual Organisation Laws of Perceptual Organisation 6. Familiarity 5.Common Fate – things can be grouped if the group appear familiar or meaningful – Things moving together in the same direction are grouped together – Eg. Look at the next picture and see if you can find all 12 human faces 9 Finding faces in a landscape Laws of Perceptual Organisation Some Additional laws of perceptual Organisation 7. Closure - Humans tend to close up a space to complete a contour (ignore gaps in the figure) Summary Laws of Perceptual Organisation Be sure you know: 8. Figure/Ground - A stimulus will be perceived how does perception happen? as separate from it's ground What is the difference between bottom up & top down processing? What do we mean by feature analysis? What does the FIT theory say? – What is the pop out effect? – What does illusory conjunction mean? What does the RBC theory say? What is the difference between the FIT & RBC theory? What are the Gestalt perceptual laws? How do these laws influence our perception? Visual Perception Part 2 In this topic, we will learn: continuation of Top-down Processing: Gestalt Approach Heuristics vs Algorithm Evidence that perception is an intelligent activity – Some behavioral and physiological evidence KMF 1023 COGNITIVE PSYCHOLOGY Lecturer: Norehan Zulkiply FSKPM Universiti Malaysia Sarawak 10  Top Down Processing: Gestalt principles Perceptual organization also occurs for hearing, speech segmentation etc. Gestalt approach applies to speech – We group or organize speech into meaningful words according to our knowledge of the language (top down perception of gestalt principles in speech) Speech Segmentation – When we listen to someone speaking in a foreign language, the words seem to be continuous, but to the Figure 3.36 (p. 81) speaker of the language who understands what the word Sound energy that results from saying the two words “Speech Segmentation.” means, the words are separate. Notice that it is difficult to tell from this record where one word ends and the other begins. (Speech signal courtesy of Lisa Saunders) – Eg. “thank you”, “terima kasih”, “xiexie”, “kamsiah”, “tochear”, “arigatogozaimas”, “merci”, “gracias denada” – Eg. “I scream, you scream, we all scream for ice cream” Gestalt principles Gestalt principles provides us with “best guess” predictions May not be accurate 100% of the time In problem solving, there are 2 approaches: 1) Heuristics – Best guess answer to a problem (shortcut approach) Figure 3.37 (p. 81) What lurks behind the – Fast, not time consuming three? – May not result in a correct solution every time What procedures 2) Algorithm (Algorithm vs. – A procedure that is guaranteed to solve a problem Heuristics) are – Takes time (STEP BY STEP PROCEDURES) involved in the perceptual system if you are to solve this problem? Heuristic for perception Figure 3.45 (p. 88) In the distance, the Cathedral of Learning on the campus of the University of Pittsburgh rises from behind a tree. Because of the occlusion heuristic, we perceive this building a continuing behind the tree. Similarly, we don’t see the person in the foreground (your author) as being chopped into little pieces by the Figure 3.38 (p. 82) fence, but we see his body It is two strangely shaped tree stumps, not an animal! as continuing behind the fence. 11  Why is perception is an intelligent activity? Main Components in Visual Perception (Human vs Computer) Gestalt principles- demonstration of how Retina Brain humans are capable of intelligent grouping of information Scene Try getting a computer to do the same Image Acquisition Image thing! Processing Light Computer 10/15/2021 Camera Why is perception so hard for computers? (The heuristics that humans use to give them the advantage over computers) Four reasons: – Stimulus received from our receptors may not be clear (ambiguous) – Objects need to be separated intelligently – Parts of an object can be hidden – Changes in lightness and darkness can be unclear Humans have built-in intelligence in our visual perceptual system, an advantage that computers do not have Figure 3.43 (p. 86) Though we may not be 100% accurate all the time… For this scene, it would be difficult for a computer to sort out which changes are due to properties of different parts of the scene and which are due to changes in illumination. Where does our perceptual Where does our perceptual intelligence comes from? intelligence comes from? Knowledge from experience (memories formed An experiment to support Neurons from birth)) specializing through experience – Experience-dependent plasticity Or Causes neurons to develop so they respond best to the types of stimulation to which the person has been exposed (bottom-up process) Neurons in our brains that respond to faces and other objects in the environment. These neurons turn the brain into a processor that is adapted to – Study with animals and humans each object that we face in the environment. – In humans, experience-dependent plasticity is Neurons specialized through: located in temporal lobe in FFA (fusiform face – Evolution – some neurons may have evolved to respond area) to specific situations/stimuli exist in the environment – FFA is activated when a person looks at faces – Experience – involve adaptation of neurons to the environment 12  Experience-dependent plasticity Experience-dependent plasticity Experiment Experiment Results (before training) – FFA neurons are activated and Experiment by Gauthier (1999) respond to human faces, less to Greebles Used FMRI to measure the activity in FFA when subjects in the experiment are shown Then subjects were given training human faces vs “Greebles” to recognise “Greebles” over 3-4 days so they became “Greebles – Greebles- computer generated images expert” Results (after training) – FFA neurons respond almost as well to human faces and Greebles With experience and new vs learning, neurons can become specialised to respond to particular stimulus => experience-dependent plasticity Where does our perceptual intelligence Summary comes from? Top Down Processing involves: – Gestalt principles of organization Brain “intelligent – Heuristics (shortcut) and algorithm approaches Knowledge stored processing” from all our 4 reasons why visual perception is hard for a computer experiences of the compared to a human: Evolution Experience- world – Stimulus received from our receptors may not be clear dependent plasticity to – Objects need to be separated intelligently (Top-down – Parts of an object can be hidden processing) respond to new experiences (eg. – Changes in lightness and darkness can be unclear Greebles experiment) There are neurons in the temporal lobe (eg. FFA area) which can be trained to specialize in perceiving human faces – evidence for process of visual perception in the (Bottom-up processing) brain Perception = TD & BU processing => Increase our chance of perceiving accurate information about our environment 13