Discussion on Global Local Image

Question: Discuss about the Global Local Image. Answer: Global-local Image An image is a visual representation of an object, an animal or a person. Images can be in the form of sculptures, paintings, photographs. They depict, give an impression, and create ideas on how things or people are. Images can be produced or made into copies and stored in electronic forms. In the reproduction of art, two general types of printable images are used: half-tone images and line art images. Solid lines describe a line art on a page that is white. Half -tone images can also be termed as photographs and are quite complicated to reproduce due to their characteristics including gray tones. Images have two major types of descriptions, global descriptions, and local depictions. It has resulted in the minting of the terminology global-local image. Images are objects and their recognition in the past recent years has been developed rapidly because of improvements in technologies like machine learning. In the modern world, new approaches have been prepared like feature extraction techniques and more image databases being available. According to Susan (2015), an images global description describes the image as a whole; they have the ability to characterize the entire objects using a single vector. The global description is defined by its texture and color (Raoui et al. 2014). The global description can also be used for classification of images. Numerous object recognition systems and machines utilize the characteristics of global features that explicitly describe a whole picture. The shape, color, and texture of a picture make them very appealing to the eye because they create very sophisticated representations of an image. These characteristics of an image are sensitive to occlusion and clutter. As a result, its speculated that an enhanced fragmentation of the picture from the background is present or an image only contains one object (Larsin, 2015). Images local description is also termed as local features of a picture. Some of the features include position which is the pixel location, scale, and orientation. In the picture below, its features can be extracted. Fig: Showing local image features An image is encoded as shown below before indicating global or local features of an image. (image from D. Lowe, CVPR03 tutorial). Advantages of encoding an image a) Locality: There is less sensitivity to the images deformations in small partial neighborhoods. Occlusion of other parts of the image can be done. b) Distinctiveness: A small false matching possibility (for example 10-3) with a 0.7 0.9 detection rate can be achieved by encoding image neighborhoods. c) Applicability: One perspective of a textured object is sufficient for training so as to identify an object from a nearby 3D viewing directions (but possibly different image locations, image orientations and different scales. Among the common key challenges in computer vision research is object recognition. The majority of image recognition systems use either of the two approaches global or local approach. This is partly because of the problems which arise when one tries to integrate a single global vector within a set of local features. In scenarios where there is rough segmentation, combining local and global features is of benefit in any application (Larsin, 2015). There are two primary methods of combining global and local elements of an image. The classical method is by stacking and the second method is using classification hierarchy. In stacking, the output of multiple component classifiers is taken into account. Using various methods for classifying has been proved to have more precise results than the component classifiers if the classifiers are accurate and diverse in nature. Classification hierarchy uses a two-tier hierarchical classification method which utilizes both local and global features in succession. At the top most position the, classes which are not separable by global characteristics are grouped together into master classes. The global feature classifier is then trained on these master classes. A local feature classifier is then trained to distinguish between the original levels contained in each super-class. When a query image is categorized to belong to the super-class, it is authorized to the local characteristic classifier, that in later ascertains to which of the categories the image will fall. The reasoning behind this is that at the top level the images are categorized into broader, more separable, groups, and at the bottom. Visual text Every image contains within it a story to be covered. The optical (visual) text refers to text in which an object performs an important function in the audiences response (Health Direct, 2014). Words and images are always combined in creative ways to create meaning and make sense. Unique features of visual language provide aid in determining the kind of message that is being conveyed by a visual text. These features include layout (color, size, shape and position), image (symbols, graphic) and writing (message, position, and font). There are many different types of visual texts namely cartoons, paintings, photographs, brochures, picture books, posters, CD covers, web pages, postcards, advertisements and film posters. Visual texts like logos, emblems, symbols and insignia communicate a meaning to us. An example of a visual text is shown below (NSW Government health. 2015). In the widely used no smoking image. Image: No smoking sign (Source: NSW Government health. 2015). The images context is in places which are designated to be no smoking zones. The image was created with the purpose of creating awareness and informs people that smoking is not allowed in those particular places. The intended audience for this particular image is individuals who smoke cigarettes and might not be aware that people are not authorized to smoke cigarettes in the marked area. The image has the intention of deterring smokers from smoking in the smoke banned zone. The symbol in the picture is that of a lit cigarette inside a red circle with a bold red backslash symbol to signify not allowed. The global factors of this image are the red color, the lit cigarette in black, the backslash sign, the smoke emanating from the cigarette, and the color white inside the circle where the cigarette is. The local factors of this picture include the color combination is red, white and black, the size of the cigarette. Others are the boldness of the backslash no smoking symbol and the fact that the image is bold red a color that is easily seen making the image stand out and be clearly even from far. Most people associate with the red color as a sign of danger and in this case scenario smokers who see the image will know that it is not allowed to smoke in the particular area (Susan, 2015). Combining images with words and other forms of communication has proved to be most effective in passing messages. The image above could be accompanied with words stating that smoking is not allowed but it is sufficient on its own to convey the message clearly without necessitating the need to use text to accompany it (Kleinberg, 2014). Apart from serving as a warning to smokers, the image above will undoubtedly act as compliance with governing laws and regulations regarding nonsmoking facilities. These pictures also convey a message to the general public notifying them that a particular location in an environment is smoke-free and conducive for the non-smokers. It is, therefore, common to find these signs and images in public recreational parks, learning institutions, bus stations, vehicles, trains, planes, offices and workplace facilities, and even in private homes and properties (Sturken, Douglas Cartwright, 2012). Health organizations also use the symbol to create and promote awareness on the hazardous effects of cigarette smoking. They do this so as to reduce the incidences of smoking as this is a predisposing factor for chronic diseases like certain types of cancer (NSW Government health. 2015.) In the onset of fast changing technology and competitive businesses, the combination of the global and local aspects of imagery will reduce errors as both provide different information about the object. If used effectively, can propel businesses and enterprises to unrivaled heights (Jensen, 2015). References Health Direct. 2014. Smoke Free England. [ONLINE] Available at:https://www.smokefreeengland.co.uk/resources/guidance-and-signage/. [Accessed 12 August 2016]. Jensen, A., 2014. Local image features. Retrieved from ftp://ftp.cs.utoronto.ca/pub/jepson/teaching/vision/2503/localFeature.pdf Kleinberg, D. 2014.Modernity and the hegemony of vision. Berkeley u.a, Univ. of California Press Larsin, D., 2015. Combining local and global image features for object class recognition. Combining local and global image features for object class recognition, [Online]. Amherst, MA 01003 USA, 1-8. Available at: https://www.google.com/url?sa=trct=jq=esrc=ssource=webcd=13cad=rjauact=8ved=0ahUKEwjA48_r37rOAhWHXBQKHZxBCP0QFghAMAwurl=http%3A%2F%2Fvis-www.cs.umass.edu%2Fpapers%2Flocal_global_workshop.pdfusg=AFQjCNFYcFFppYTl8pbgc9wtjYFN8O3qqwsig2=vfYhIRtU4XoIfKII6MIr8Q [Accessed 12 August 2016]. Margaret R., 2016. Image. [ONLINE] Available at: https://whatis.techtarget.com/definition/image. [Accessed 11 August 2016]. Moriarty Mitchell Wells Crawford Brennan Spence-stone, S. N. D. W. D. R. L. R., 2015. Advertising: Principles and practice. 3rd ed. pg. 502: Pearson Australia Group Pty Ltd. Marita S. Lisa C., 2013. Practices of looking: an introduction to visual culture. NSW Government health. 2015. Smoke-free signage and resources. [ONLINE] Available at: https://www.health.nsw.gov.au/tobacco/Pages/smoke-free-resources.aspx. [Accessed 11 August 2016]. Raoui, Bouyakhf, Devy, Regragui, Y. E. M. F., 2014. Global and local image descriptors for content based image retrieval and object recognition. Applied mathematical sciences, vol. 5, 2011, no. 42, 2109 - 2136. Susan, M. 2015. Seeing Global. From: https://susanbuckmorss.info/constellation/seeing-global/ Sturken, M., Douglas, S. J., Cartwright, L. (2012).Practices of looking: an introduction to visual culture. Don Mills, Ont, Oxford.