Please read chapter 4. After reading chapter 4, please respond to the following questions:
What were two things from the chapter that you found interesting? Why were they interesting to you? Which two things did you find the least interesting? Why? What did you read in the chapter that you think will be most useful to in understanding the visual system? Why? Finally indicate two topics or concepts that you would like me to cover in more depth in class.
Note: Keep in mind that you will be allowed to bring in the blog posts to class with you when you take exams. Be sure to use the terms and terminology in your posts.
Let me know if you have any questions,
--Dr. M
The first thing that interested me was the broad topic of middle (midlevel) vision which is a stage of visual processing that comes after basic features have been extracted from the image (early vision) and before object recognition and scene understating (high level vision). The act of recognition must involve matching what we perceive now to a memory of something we perceived in the past. The goal of middle vision is to organize the elements of visual scene into groups that we can then recognize as objects. Finding edges is a good starting place to recognizing objects. There is a thing called illusory contour which is perceived, even though nothing changes from one side of the contour to the other in the image, for example a white arrow on a white background. Furthermore, lower level visual processes deliver fairly straightforward bits of information about a line here and a color there. Information needs to be understood before we know what we are seeing. Middle vision behaves like a collection of specialists, each with a specific area of expertise and individual opinions about what the input might mean.
The second thing I found interesting was that of object recognition which is one of the key tasks of visual perception. A detailed point by point representation of objects is not transmitted to the parts of the brain interested in objects. The raw image is analyzed by the middle vision process and it is the output of those processes that will be available to object recognition processes. Since this was such a short section I also found faces fascinating. Prosopagnosia is the inability to recognize faces which I didn’t think could be that hard until I got tricked in the book under figure 4.34. If an individual has damage to their brain they could face prosopagnosia and double dissociation. Double dissociation is a phenomenon in which one of two functions, such as hearing and sight, can be damaged without harm to the other, and vice versa. For example, a person who is blind may still hear or be deaf and still see. Also, in face recognition, it is possible to lose the subordinate ability to recognize specific faces while retaining the ability to recognize an object as a face.
One thing I found least interesting was texture segmentation and grouping which deals with connecting little pieces of line segments only as far as dividing the raw image into objects. Texture segmentation is carving an image into regions of common texture properties. In other words, the portion of the image with a coarser texture is separated from the rest of the image. In addition, texture segmentation is closely related to similarity and proximity. Similarity is a Gestalt grouping rule stating that the tendency of two features to group together will increase as the similarity between them increases. Whereas, proximity states that the tendency of two features to group together will increase as the distance between them decreases. Texture grouping can be based on similarity in a limited number of features such as color, size, orientation, and aspects of form. Also, combinations of textures do not work well together. There is also parallelism stating parallel contours likely belong to the same figure and symmetry stating that symmetrical regions are likely to be seen as figure. There are two other Gestalt grouping rules which are common region and connectedness meaning two feature will group together if they appear to be part of the same larger region and two items will group together if they are connected. Camouflage is cool though, especially how animals’ colors help them hide from predators.
A second topic that wasn’t as interesting was templates versus structure. Geons which are perceptual objects that are built. Geons are specified as collections of nonaccidental features, therefore, a visual system should be able to recognize a geon equally accurately and quickly, regardless of how the geon is oriented in space. Objects are recognized by the identities and relationships of their component parts known as the recognition by components model. Structural descriptions composed of geons and relations should be viewed point invariant. One meaning a property of an object that doesn’t change when the observer viewpoint changes and secondly, a class of theories of object recognition that proposes representations of objects that don’t change when viewpoint changes. In addition, there is structural description which is a description of an object in terms of the nature of its constituent parts and the relationships between those parts. For example, the image of the letter “A” is being matched to a structural description of an “A.” The idea is that we recognize objects by matching pixels or every low level feature of the input to a representation in memory also known as naïve template theory which proposes that the visual system recognizes objects by matching the neural representation of the image with a stored representation of the same shape in the brain.
I think figure and ground is useful in understanding the visual system because it is the process of determining that some regions of an image belong to a foreground object and the regions are part of the background. A famous figure-ground figure is the vase/face which looks like two people’s side profile or a vase in color. A rule for figure-ground assignment is surroundedness which stats that if one region is entirely surrounded by another, it is likely that the surrounded region is the figure. There are a few other principles when it comes to regions to figure ground assignment such as size, symmetry, parallelism, extremal edges, relative motion and surroundedness. In addition I like the Gestalt theory which holds that the perceptual whole is more than the sum of is sensory parts and the Gestalt grouping role guides the visual system in its interpretation of the raw retinal image.
Topic/concepts I would like covered more in depth would be objects in the brain, Necker cubes and occlusions. Occlusions show that if something stops then there must be something else getting in the way or hiding it from view. The visual system seems to come up with the reason that there is another contour occluding the vertical line, with the occluding edge oriented perpendicularly to the occluded edge. Occlusion deals with relatablity which is the degree to which two line segments appear to be part of the same contour. A Necker cube is an outline that is perceptually bi-stable. Unlike the situation with most stimuli, two interpretations continually battle for perceptual dominance. Every image is, in theory ambiguous, but the perceptual committees almost always agree on a single interpretation. Moreover, objects in the brain involve the extrastriate cortex which is the region of the cortex bordering the primary visual cortex and containing multiple areas involved in visual processing. Visual pathways seem to be important in processing information relating to the location of objects in space and the actions required to interact with them.
I also agree that if we have to chance to go more in-depth about the extrastriate cortex it will help us processing and understand the information, relate and interact with the objects. Can't wait to hear about it.
One topic that I found interesting was the idea of Perceptual Committees. I found this interesting because our eyes collaborate on what the have seen. Sometimes these ideas collide and the results can be unpredictable. However, a consensus view almost always rapidly emerges and our eyes settle on a single interpretation of the visual scene, but these decisions do not need to be final. Sometimes, an ambiguous figure will generate more two or more plausible interpretations. However, the perceptual committee is unwilling to readily consider any of the infinite possible interpretations. This comes about because of accidental viewpoint or the precise location were you can see something. Often moving slightly one way will make you unable to see the object in that way, so our perceptual committees know this and know not to place what they see on them. Another topic that I found interesting was figure and ground. Figure and ground much also operates using a perceptual committee to come to an agreement on how images should be understood; these committees are governed by a collection of principles acting together. In certain cases the perceptual committee may have a difficult time reaching a consensus because all the visual stimuli may be ambiguous, but the committee almost always comes up with a single conclusion. There are 6 principles at work here: surroundedness, if one region is entirely surround by another it is likely that the surround region is the figure, size the smaller region is likely to be the figure, symmetry a symmetrical region is more likely to be seen as a figure, parallelism regions with parallel contours are more likely to be seen as the figure, external edges, and relative motion. One topic I didn’t find very interesting was the area on faces, because while an interesting topic I felt that there was limited information in the book. Another area I didn’t not find very interesting was texture segmentation and grouping, I found this topic fairly difficult to understand. I think that understanding perceptual committees and pattern recognition will be key to understanding the visual system. Two topics I would like to go over in class are middle vision and object recognition.
Some more info on object recognition from chapter 4
Naïve template theory is the proposal that the visual system recognizes objects by matching the neural representation of the image with a stored representation of the shape in the brain. This system operates like a lock and key. Structural descriptions play an important role in this theory, they are description of an object in terms of the nature of its constituent parts and the relationships between those parts. Another theory associated with object recognition is recognition by components or the idea that objects are recognized by the identities and relationships of their component parts
I thought the topic of middle vision was also pretty interesting. Middle vision i when we organize the elements of a scene into groups that we can make into objects. The first step of middle vision would be finding edges. Neurons in the striate cortex can find and detect parts of lines. However, we need to be able to decipher which lines belong to which objects. In fact, we can sometimes perceive lines that are not really there. One example was the picture of the house in the book. The circle objects with bits cut out of them created the illusion that the house had its own outline. These are called illusory contours. Structuralists such as Wundt and Titchener stated that perceptions are made up of atoms of sensation. This includes bits of color, orientation, etc. The illusory contours go against this view because there are no lines or physical edges that create this image. The Gestalt grouping rules are a set of rules that state which parts of a picture will seem to be grouped together. These lines seem to support each other creating the image that they are one. Good continuation is also when two parts or elements of a picture seem to lie on the same contour, which usually leads us to see them grouped together. I thought all of this was very interesting, because I never thought of things in that sort of way.
I thought the topic of Texture Segmentation and Grouping was very intersting as well. It tied in with middle vision and finding edges. Texture segmentation is when we find an image in regions of common texture properties. There are two strong principles in texture segmentation. These are similarity and proximity. Similarity is when the similar pieces in the picture with similar texture will be grouped together. The tendency to group features together increases as similarity increases. Some examples may be color or size. Proximity is when the tendency to group objects together increases as the distance between them decreases. So items that are close to each other are more likely to be grouped together than those that are far away. Parallelism and symmetry are also principles for this. Symmetrical figures create an image. I thought these things were very interesting as well.
I think there were certain topics in the Object Recognition category that weren't quite as interesting. For example, the naive template theory is the idea that we are able to recognize certain objects by matching the pixels or matching a low-level image with a picture in our memory. Locks and Keys were a big part of this theory. The example used was the A. The A is put into the blocks , and since it is the same shape as the A to the side of it, it is seen as an A template. There is a problem with this theory however, and that is that there are too many templates required. Although this topic was interesting as well, it took a little more reading and concentration to catch on to the general idea. Structural description was also one of the less interesting things to me. This is a specification of a picture or object in terms of that object's parts and the relationship of those parts. This includes geons, or components of which objects are built.
I found the topic of objects in the brain to be very difficult to grasp and remember (such as the extrastriate regions, inferotemporal regions, and other things). It is hard for me to grasp those things as a whole. I would also like to learn more about Middle Vision that I may have missed, because I found that to be the most interesting. I think middle vision may be most helpful in understanding our vision and how we perceive things as well.
I'd like to add some other topics about Object Recognition that interested me as well. Along with object recognition The geons go along with the "recognition by components" model which is a model of object recognition stating that objects are recognized by the relationships of their components. Viewpoint invariance, however, is described as a property of an object that doesn't change when the viewpoint of the observer changes.
There is also the idea that there are multiple recognition committees. This is the idea that recognition may not be one single act. Instead, we can see and recognize certain objects in many ways. For example, when we see a bird, the entry-level category is the label that comes to mind most quickly. There might be different subgroups of these entry-level categories, but the general name is bird.
I think the most important part of this chapter to understand the visual system is the discussion of middle vision. I also found the most interesting things within this discussion were the various principles that work together in middle vision. According to our book middle vision is a stage of visual processing. This stages if after early vision which processes basic features of an image and before high-level vision which includes object recognition and scene understanding. Middle vision organizes the various elements of a scene into groups so that it can be recognized as objects. Middle vision does this through a series of interesting processes including: finding edges, using Gestalt principles, and occlusion. Finding edges along with the edge quality is important to begin the process of recognizing an object. The mind can perceive edges even when there is a lack of physical evidence for them, called an illusory contour.
The topics I found interesting were the discussion of Gestalt principles and occusion. Gestalt (perception) psychologists Wertheimer, Kohler, and Koffka defined a series of principles that allow our visual system to process and recognize objects. Gestalt grouping rules refer to a set of rules that describe which elements will group together. One rule is the principle of good continuation which states that two elements will group together if the lines have the same orientation (lie on the same contour). The similarity principle states that as the similarity between two elements increases, the tendency for them to be grouped also increases. Similarity can include: color, size, orientation, etc. The proximity principle states that the closer items are together the more likely they will be grouped. The parallelism principle states that elements that are parallel to each other are likely to be grouped. The symmetry principle states that elements that are symmetrical to each other are more likely to be grouped. The common region principle states that if two elements are a part of the same larger region they will be grouped together. The connectedness principle states that two elements will be grouped together if they are connected. Figure-ground assignment is an important part of middle vision that determines the image of the foreground (figure) from the background (ground). Figure-ground works off of its own principles including: surroundedness, size, symmetry, parallelism, extremal edges, and relative motion. These rules help the middle vision process an image for object recognition. Occlusion is also an interesting part of middle vision. Occlusion occurs when one contour occludes or obstructs another. Our visual system allows us to recognize that the contour would continue without the occluding contour.
The first topic I found less interesting was the idea of perceptual committees. This is the idea that our visual systems “behaves like a collection of specialists” where they discuss their opinions on what the visual system is trying to perceive and recognize. Physically this committee probably consists of interconnected set of neurons taking in input and producing output. The perceptual committee has two rules: to honor physics, and to avoid accidents. The committee recognizes and often ignores accidents via an accidental viewpoint, a viewing position that produces regulatory in the image that is not in the real world. Another topic I found less interesting was the discussion of object recognition theories. The first theory is naïve template theory and it states the visual system matches every low-level feature with a stored representation in the brain. However, if this theory were true it would take a lot of our brain capacity to develop a template for every object. A second theory is the structural description, in terms of its parts and the relationships between those parts, determines our recognition of objects.
Two topics I would like to discuss in class are facial recognition the discussion of “Objects in the Brain” on pgs. 97-101. I think they are both interesting but also a little more complex so further discussion would help in my understanding of the topics.
Chapter 4 Perceiving and recognizing objects was very interesting. The best part of it was Middle Vision which is a loosely defined stage of visual processin that comes after basic features have been extracted from the image. Middle Vision helps to organize the elements of a visual scene into groups that we will be able to recognize the objects. Finding the edges is important and finding gaps that we do not see. The author gives couple examples of it; houses and elephanes.
Another thing that was interesting is object recognition and specifically recognize faces. I liked the images of two faces that look exactly the same but when we turn the faces right side up one of the pictures will look wrong. It is a different level of categorization. The process that recognize face as a face care little about inversion.
Probably the least interesting thing in this chapter was about recognision by components model and viepoint invariance. The first one is Biederman's model of object recognision which the object is recognized by identities and relationship of their componenet parts. And viepoint invariance an object doesnt change when observer view poin chages. Somehow it is interesting but it is quite obvious vthat if we combine some things or addes to some other objects we can see it quite differently or even see something else; new figure.
Most useful would be in understanding visdual system will be part objects in the brain, how does that work, how the visual system might go about recognzing objects to physiological studies of actual visual systems.
I would like to learn more about prosopagnosia- disorder which a person can not identify faces.
I found most of the chapter interesting but two concepts I found the most interesting was prosopagnosia and the perceptual committees. Prosopagnosia is the inability to recognize faces. This is caused by damages to certain parts in the temporal lobe of the brain. I find this so interesting because even with this condition, a person can still recognize that the object in front of them is a face. There’s another similar disorder called agnosia that is the inability to recognize objects, but they can also still recognize that something is in front of them. This proves that there are different parts of the brain that recognize different things.
The other concept I found interesting is the idea of a perceptual committee. Of course, the brain doesn’t have an actual group of people that control how we perceive things, but the idea of having rules and principles about how we see things results in that kind of idea. This committee decides how we interpret the things we see. Sometimes, we get confused by illusory contours like the Necker cube, which confuses us because we’re not sure on what type of shape to think it is.
Two things I found least interesting is structuralism and lesion. I think structuralism is uninteresting because I’ve already learned about it in history and systems and I don’t find the concepts interesting either. I think that lesions are boring also because they don’t have to do with the visual system, just how doctors can get into the brain.
I think illusory contours help to understand the visual system because even though they trick the visual system, it explains why our visual systems work the way they do and why they can be tricked. Some concepts that I’d like more information about are viewpoint invariance and accidental viewpoint.
Lesions of the brain allow us to know what parts of the brain are responsible for what functions. I think the authors were trying to explain why we know certain things about the visual system based on the discoveries due to brain lesions of monkeys. Through such studies they learned that damage to the temporal lobe (thus lesion of the temporal lobe) resulted in agnosia as you described above. Lesions of the temporal lobe have a lot to do with our understanding of the visual system.
After reading chapter four, there were a few things that stood out to me. The first thing that I found interesting was the basic concept of middle vision. As defined by our text middle vision is a stage of visual processing that comes after basic features have been extracted from the image ("early vision") and before object recognition and scene understanding (high-level vision). The main goal of middle vision is to organize the elements of a visual scene into groups that we can then go on to recognize as objects. The book points out that the occasional lack of edge does not bother our visual system, and that it will just fill in the missing information. For example, the text shows a picture of an arrow which varies from light to dark, placed on a background that also varies from light to dark. there are points at which the degree of color is the same on the arrow as it is on the background but our visual system is still able to detect the image. This is an example of an illusory contour (a contour that is perceived, even thought nothing changes from one side of the contour to the other in the image).
Another topic that I found interesting in chapter four was figure-ground assignment. Figure-ground assignment is defined in our book as the process of determining that some regions of an image belong to a foreground object (figure) and that other regions are part of the background (ground). To determine what in an image is the figure and what is the ground, it is important to understand the concepts of parallelism, and symmetry. Parallelism is a rule for figure-ground assignment stating that parallel contours are likely to belong to the same figure. Symmetry is also a rule for figure-ground assignment stating that symmetrical regions are more likely to be seen as a figure. An example of figure-ground assignment is the visual perception of the classic vase/face figure. Normally the perceptual committee easily reaches a consensus as to what it is seeing. When visual stimuli may be ambiguous, the processes that determine figure and ground almost always manage to come to a single conclusion and we are surprised when we see something else. Other concepts that are important to keep in mind when determining figure-ground assignment is the concept of surroundedness which is a rule for figure-ground assignment stating that if one region is entirely surrounded by another, it is likely that the surrounded region is the figure. The other concept to keep in mind is size. The smaller region is likely to be the figure.
One of the things that I found least interesting was the global superiority effect. This is defined in our text as the finding in various experiments that the properties of the whole object take precedence over the properties of parts of the object. This basically states that the first goal is to carve the retinal image into large-scale objects. Since most objects are composed of parts, we can then recognize the individual parts after we recognize the large-scale object.
Another thing that I didn't find very interesting in this chapter is the idea of non-accidental features. According to the text, non-accidental features are features of an object that is not dependent on the exact (or accidental) viewing position of the observer.
I think that the topics of templates and structural descriptions are important to understanding how the visual system works. Template theory and structural description are both theories as to how we recognize objects. Template theory is the proposal that the visual system recognizes objects by matching the neural representation of the image with a stored representation of the same shape in the brain. The problem with this theory is that you would have to have way to many templates in your brain than your brain can actually handle.
A structural description is a description of an object in terms of the nature of its constituent parts and the relationships between those parts. One example of structural description is by taking geons (geometric ions out of which perceptual objects are built) and building objects and then remembering what parts were used to build it. this is known as recognition by components model (Biederman's model of object recognition, which holds that objects are recognized by the identities and relationships of their component parts). One of the problems with structural description theories is that they tend to be very viewpoint-dependent. Also, it is not clear that geons or any of the other "alphabets proposed by structural description models are adequate for the language of object recognition.
There are a few topics that I would like to go over in class to gain a better understanding. I would like to talk more about the where pathway and the what pathway. The book states that the visual areas in the where pathway seem to be important in processing information relating to the location of objects in space and the actions required to interact with them. The where pathway also plays an important role in the deployment of attention. the what pathway appears to be the locus for the explicit acts of object recognition.
I found middle vision to be interesting. Middle vision is the processing between the basic vision but before object recognition and interpretation (high level vision). The ultimate goal of middle vision is to group items so that they can be recognized as objects. The five principles of middle vision were putting together what goes together, separating objects from the background, dividing objects based on knowledge of image formation, avoid interpretations that require accidents, and seek consensus while avoiding ambiguity. I found the importance of point of view of the objet to be very interesting. Upright faces are more easily recognize than faces that are upside down. Apparently there are also specific parts of the brain for faces I would like to learn more about this in class as well as prosopagnosia. This is the first chapter that I found it difficult to find topics that I liked least but I supposed I would have to say the section on occlusion was the least interesting to me but only relative to the other topics. Naïve template theory was also less interesting to me because I found the section to be a little dry.
I found middle vision to be interesting. Middle vision is the processing between the basic vision but before object recognition and interpretation (high level vision). The ultimate goal of middle vision is to group items so that they can be recognized as objects. The five principles of middle vision were putting together what goes together, separating objects from the background, dividing objects based on knowledge of image formation, avoid interpretations that require accidents, and seek consensus while avoiding ambiguity. I found the importance of point of view of the objet to be very interesting. Upright faces are more easily recognize than faces that are upside down. Apparently there are also specific parts of the brain for faces I would like to learn more about this in class as well as prosopagnosia. This is the first chapter that I found it difficult to find topics that I liked least but I supposed I would have to say the section on occlusion was the least interesting to me but only relative to the other topics. Naïve template theory was also less interesting to me because I found the section to be a little dry.
One thing that I found interesting about the chapter was the section on finding edges under the middle vision section. Middle visions goal is to organize the parts of a scene into groups so that we can then recognize them as objects. The portion of this that I found interesting was about finding edges. The text book had a picture of an arrow that was shaded in black/gray that was in a box that had the same shading. When looking at the picture you can definitely see the outline of an arrow. But the text points out that some of the edges are not there, because the color of the background and the color of the arrow are the same. Our mind overlooks this and fills in the edges to fit with out picture of the arrow. This is called illusory contours.
Another portion of the chapter that I found interesting was the information about agnosia. People with agnosia have the ability to see but are not able to know what they’re seeing. From what research shows this is based in the temporal lobe, researchers were able to use stroke victims who had lesions in the temporal lobe to do their studies. Agnosia can be very specific. One type is prosopagnosia. This is the ability to recognize a face as a face but not recognize who the face belongs to. Others include the abiity to recognize animate objects but not inanimate objects.
There wasn’t anything in the chapter that I found uninteresting. However, there are a couple things I feel like I need to get a better grasp on. I would like to learn more about the section on templates versus structural descriptions. This section covers naïve template theory and then spoke about geons. I don’t feel that I fully grasped the information so I would like to cover more of this.
The other section that I would like to learn more about is the objects in the brain. For some reason I can’t seem to get a grasp on the information. This section covers the parts of the brain that are involved in things such as grouping and texture segmentation. It also covers the what and where pathways. I believe going over this a little more will help in furthering my understanding of the chapter.
I believe the most helpful part of the chapter was the whole section on middle vision. Again, middle vision’s goal is to organize parts of a scene into groups so that we can recognize what we are seeing. Learning how we do this through finding edges, looking at texture and groupings, and occlusion makes me look at things that I see a little differently.
The two most interesting things that I read in chapter 4 were illusory contour and prosopagnosia. Prosopagnosia was chosen because, well it’s just cool that you can maintain the ability to see a face while losing the ability to recognize who the faces belong to. I can only imagine the frustration such an injury would inflict. As for illusory contour, I chose this because it was the first thing that got my attention and I perked up to learn more about it. I find it interesting how our brains interpret pretty straightforward information and make it more complex, seeing patterns where technically non exist, seeing a shape rather vividly when it is only a suggested outline.
Finding two things in chapter 4 that I was less than interested in was difficult, I suppose that is because I tend to skim over things that are of little interest to me, but also because this chapter is chalk full of information that was interesting. Since I have to choose something I am going to go with the heuristics used in vision, even though I find that too interesting, it is a concept that I have learned previously and thus is lacking in novelty.
I would like to learn more about naive template theory in class, also toward the end of the chapter they explain more about where these processes of recognition are taking place in the brain, I would like to hear more about that.
Of course understanding middle vision will be the most helpful, the next step in the staircase of the visual process, with it's goal of "organizing the elements of a visual scene into groups that we can recognize as objects."
One topic that I found very interesting was the gestalt principles. I love pictures and seeing how we perceive that picture a certain way opposed to another. I also like the texture pictures because I never notice how I don't pay attention to the colors just the different shapes. I also found it interesting that cells will fire to certain stimuli like the monkey example.
Two things I found that weren't very interesting were templates vs structural descriptions especially geons. I thought it was just a bunch of words and wasn't as interesting as what I had read earlier in this chapter.
This helped me understand how the visual system works because I'm learning how my visual system sees and perceives certain things and why. I go through the pictures and can say oh yeah I definitely do that.
Two things I would like to hear more about are facial recognition and objects in the brain and how it processes the info
One of the topics from chapter 4 that I found interesting was the Gestalt grouping rules. Similarity, Proximity, Good Continuation, Common Region, Connectedness, Parallelism and Symmetry are crazy to think about when you start to compare them to things in your everyday environment. The similarity principle can be used in almost anything and we are taught to notice similarity from a young age. Proximity is like finding a mate...when you spend time together you start dating...haha. Good Continuation is when two elements will tend to group together if they seem to lie on the same contour. Common Region is a rule stating that two features will tend to group together if they appear to be part of the same larger region. Connectedness is a grouping rule stating that two items will tend to group together if they are connected. Parallelism is a rule for figure–ground assignment stating that parallel contours are likely to belong to the same figure. Parallelism is my favorite because it reminds me of the constellations. And last but not least, Symmetry is similar to parallelism. Symmetry is a rule for figure–ground assignment stating that symmetrical regions are more likely to be seen as figure.
My second favorite topic was Naïve template theory. The naïve template theory is the proposal that the visual system recognizes objects by matching the neural representation of the image.
The idea above is a lock and key template. A letter A stimulus falls on an array of spot detectors. If the A falls on the filled detectors and not on the empty ones, it is identified as an A. Therefore, the array detectors serves as an A template. The down fall is that letters are very common structures. Our brain will automatically read the common shape.
One thing that I found least interesting … I really like this chapter but if I had to choose it would be the section about the faces. It talks about how faces are easier to recognize upright than inverted. That is because we don’t normally see faces upside down, backwards or arranged differently very often so it is something new for us to process.
The second thing I find less interesting is the section on ambiguous figures. Ambiguous figures are a visual stimulus that gives rise to two or more interpretations of its identity or structure. Committee Rules: Honor Physics.
The most useful information would have to be the extrastriate cortex which is the region of cortex bordering the primary visual cortex and containing multiple areas involved in visual processing. The extrastriate cortex is the locus of mid-level vision. Neurons in the extrastriate cortex generally respond to visual stimuli within their receptive fields. These responses are modulated by extraretinal effects, like attention, working memory, and reward expectation.
Finally, the two topic that I would like for you to cover more in depth in class is the extrastriate cortex which is the region of cortex bordering the primary visual cortex and containing multiple areas involved in visual processing. The extrastriate cortex is the locus of mid-level vision. Neurons in the extrastriate cortex generally respond to visual stimuli within their receptive fields. These responses are modulated by extraretinal effects, like attention, working memory, and reward expectation. What if that part of your brain was damaged, how much processing would you lose? An ambiguous figures…Maybe you can make it more interesting.
Cognition, in many respects, is just a filter of information. Our brain seeks to process what it has to with as little energy spent and as little cognitive effort put forth as possible. Beginning at the cellular level on the retina, our bodies are tuned to pick up specific information. The organisation of ganglion cells through the striate cortex organizes the information being processed and looks for specific cues. Straight lines and corners are cues because they hold information. Cortical cells in the striate cortex organize the information and begin to form the image we see in our mind by recognizing and recreating lines and corners. The next step in the process is to recognize what the lines and corners mean. What patterns do they form and what do these patterns mean? Object recognition is a fundamental part of how we visually experience the world. Gestalt psychology was a movement that took advantage of object recognition to promote it's theory. Gestalt holds that the whole is greater than the sum of its parts. Basically our brains operate in such a way that transcends analyzing every small detail of objects and adding them together to get all the information we have about said object. Rather our brain naturally looks for a pattern, and once the cues have been established to show that the pattern exists our brain infers a lot of information from a preconceived schema. There are a handful of gestalt principles that seem to hold true for most everybody involving properties of objects and how our brains tend to process them. For example the principle of proximity states that items that are close together in distance are likely to be grouped, as opposed to object which are further apart. These principles become pretty intuitive when you observe the patterns and pictures used to illustrate them.
Figure and ground is a concept that is illustrated in many gestalt style patterns and is important in our focused vision. Basically everytime we are focused on an image our brains assign a figure and a ground. Essentially the figure is the subject, the lady on stage. Our brain uses 4 principles to assign the role of figure. The first is surroundedness and refers to usual tendency for the ground to completely or almost completely surround the figure. Size is another factor. Using knowledge of the normal world our brain reasons that the figure will be smaller than the ground, the book uses the example of a cow being much smaller than the field it is standing in. The third factor our brain looks for is symmetry, our brain is more likely to assign to figure an object that displays symmetry. And the fourth dimension is parallelism refering to our brain's tendency to pick out parallel objects as figure.
I will post more on this chapter as well I have only gotten to read a little more than half but im so tired. I would like to hear about some of the physiological stuff behind object recognition, the section in the book about cells tuned to faces looks interesting although I haven't gotten there yet. I love the cognitive aspect as well I am a fan of endless speculations so It would be great to hear a little of that in lecture as well.
The concept I find very interesting in chapter for is Gestalt Psychology. Its most interesting to me because it shows up so much in our culture. Advertisers, artists, and visual designers understand the principles of Gestalt and it influences the images we see. Gestalt is a German derived school of thought that translates to the word 'form'. The main tenant being that the perceptual whole could be greater than the apparent sum of all the parts. Gestalt researchers compiled 'grouping rules' that show how the brain interprets the raw retinal image. These rules are: similarity, continuation, orientation, proximity, region, parallelism, and connectedness. These rules show that even when a pattern doesn't inherently exist, your eyes will try to make a pattern out of what is being seen. This is done through perceptual committees that act like a group of specialists in that they quickly come up with a single answer from the several possible answers that could be derived from one image.
This information is useful to me because understanding how I perceive the world is incredibly interesting, and now when I see the cliche' vase/face figure I can explain why that happens.
The topic I would like to be covered in class is a basic outline of middle vision to "fill in" what I've missed.
Chapter 4 was a very interesting one. Two things that I found interesting are figure-ground assignment, and the part about faces. First, I found the figure in the ground assignment to be very interesting because it is something that made me really think. We too often assume that something is the figure and something is the ground (background). The blue and green picture in the book pointed out that we aren’t always right. Figure and the ground is Gestalt psychology. The example in the book is the picture that looks like both a vase and two faces. These type of pictures have always interested me. The second thing I found to be very interesting was the part about faces. I thought prosopagnosia was very interesting. I cannot image going through life while suffering from this. Prosopagnosia is the inability to recognize faces. This ties in perfectly to Agnosia. Agnoisa means that you can see something but you do not know what you are seeing (ex. faces). This can be due to an accident or trauma. The part of the brain that is often affected is the inferotempral (IT) cortex. This is in the cerebral cortex.
The thing that I found least interesting was texture segmentation. I found this to be not interesting because although it told us about texture, I struggled to find what it was trying to say about vision. Similarity tells us that similar textures are going to group together because of similar properties. I just feel like I got nothing that helps me connect texture to the overall picture of how we perceive and recognize objects.
Other than that I honestly found the chapter to be really interesting. I loved the it walked us through the middle vision and used the analogy of committees. It was interesting how throughout the chapter each “player” in the committee played its own very important role and how many things went on at the same time. I also loved the figure and the ground stuff. This is stuff that seems so elemtray, yet we so easily look over it. I love the example in the book of the blue and green picture that made us question, what is the background. I also found heuristic to be very exciting. Heuristic is a mental shortcut; something we don’t even contiously think about. I thought parts and whole were very interesting. The global superiority effect, to me was another thing that we othen overlook. This stated that there are smaller things within bigger pictures (ex. big letters made out of little letter). The template theory was something that helped me better understand what goes on in the middle part of vision. This theory states that our visual system recognizes the shape of something and stores it in our heads.
Overall, the thing that I read in the chapter that will be the most helpful in understanding the visual systems was the committee analogy. To me this explains the complexity of what is really going on in a much simpler way. There are many different processes at work at the same time. We know a lot about the different systems and basic principles of object recognition but it is very complex.
The two things I would like for you to cover more in class are the Gestalt theory and occlusion. I feel like both of those topics were just not defined very well in the book.
Just to catch up on what I have leanred before the test...
The purpose of the visual system is to interpret images. The structure parts of an eye are, Cornea and Lens, Iris, Retina, and LGN. The purpose of the cornea and lens are, the cornea bends light (it is transparent), lens fine tunes it. In other words it adjusts and focuses light. The job of the iris (the color part of the eye) is to regulate the amount of light. Next, the retina is at the back of the eye, it receives an image and sends it to the brain via the optic nerve. It is sent to the LGN. Another important part of the visual system is the striate cortex. This analyzes LGN inputs.
Chapter 4 talks a lot about middle vision. The purpose of middle vision is to combine features into objects. Another name for middle vision is midlevel vision. Brains job to create a stable predicable world The book describes middle vision as a loosely defined stage of visual processing after early vision but before high level vision. The purpose of high level vision is to recognize objects. The way middle vision works is by putting things into groups so that our brain can recognize them as objects. One very important part of middle vision is finding edges. In class we have talks about the importance of edges many times. The example of finding edges in the book is a house in which some lines are taken away but all the edges are still there. Our brain is still capable of putting together that the picture is still a house. Illusory contours are also important in middle vision. Illusory contours is something that is perceived even though nothing changes. Moreover, structuralism is also important in middle vision. Structuralism is the thought that complex objects could be understood by analyzing components. Gestalt played a key role in rules that make contours. Gestalt theory says that the whole is greater than the sum of its parts. Gestalt grouping rules are how the Gestalt theory is explained. Gestalt perceptual groupings include: similarity, proximity, parallelism. These are a set of rules that tell which parts of a picture will group together. If two elements will likely group together, this is called good continuation. An example from the book that we also saw in class about illusory contours is the house that looks like it has pac man circles on the edges. Occlusion plays a prevalent role in this chapter. Illusory contour is our brains filling in lines. Another example of this is class is the guy sitting in the plumbing truck, half of his body was painted on the outside. Other jobs of middle vision are Group and texture segmentation, figure-ground assignment, and edges, parts, and wholes. Simliarity, proximity, figure and ground, and surroudedness are all very important in texture segmentation and grouping. Scene segregation also has to do with texture. This is because we tell objects from the background by way of texture and color. The example given in class was the picture of the dog that you showed us. Here is a quick summary of the goals of middle vision: 1. Bring together what should be brought together. 2. Split what should be split. this is where we find edges. Use what you know. 4. Avoid accidents. 5. Seek consensus and avoid ambiguity.
Next, heuristics are a big section of this chapter. A heuristic is a mental shortcut. Some heuristics that our brain uses are Arrow junction, Y junction, and T junction. Occlusion plays a significant role when talking about heuristics. Something that I found to be very interesting was what David Navon figured out. The exercise in the book was the one with the big letter, first he told us to glance and see if we can spot an “H”, next he told us to glance and see if we could find a “u”. It took longer to find the smaller “u”. This is an example of the global superiority effect. The global superiority effect states that the entire object is more important to our brain than the parts that make up the whole object. Other heuristics the brain uses include, nonaccidental features, parts and whole, object recognition, template theory, structural description, geons, recognitions by components model, and viewpoint-invarient. First, nonaccidental features are features on objects that do not depend on where the observer is viewing it from. Second, when talking about parts and wholes, this is where the global superiority effect falls into place. Third, heuristics are directly related with object recognition. Object recognition is matching of what you see with memory. The example given in the book is if you see a picture of the president, you might be able to store it in your brain with a “president” node. Fourth, template theory states that our visual system recognizes objects by matching the neural representation with stored “shapes” in the brain. The example in the book for a basic template is the letter “A” with some missing parts, but basically the same shape. Next, structural description is a talking about an object in terms of its parts and the relationship among those parts. Next, geons are nonaccidental features. Next, we must talk about the recognition by components model. This is where geons come into play. This model was proposed by Biederman. This model states that objects are recognized by the identities and relationships of their component parts.
Back to object recognition. Viewpoint-invariance is worth mentioning. This is when the properties of an object stay the same no matter the viewpoint of the observer. Moreover, there are also such things as multiple recognition comities. The reason for this is because we may perceive an object multiple different ways. The example given to us in the book for the need of multiple recognition committees is the picture with for birds. They are all birds, but they are all different. This section of the chapter also talks about prosopagnosia. This is an inability to recognize faces. Next, extrastriate cortex plays a very import part in object recognition. This is a part in the brain that contains areas that are responsible for visual processing. Moreover, infertemporal (IT) cortex is also very important when talking about object recognition. This is a part of the cerebral cortex in the lower temporal lobe that is important in object recognition.
The brain resolves ambiguity by heuristics. This again is a mental shortcut. This works most fo the time but not all the time. “rule of thumb”. The difference between heuristics and algorithm is that algorithms always work.