Princeton COS 436 - Sonorific Toaster: Image Processing

Unformatted text preview:

1/13/09 6:16 PMSonorific ToasterPage 1 of 10http://www.princeton.edu/~malia/report.htmSonorific Toaster: Image Processing COS436/ELE436: HCI Technology Professor Perry CookJanuary 13, 2009SchGaDY's Musical Toaster Malia Douglas, Dmitri Garbuzov, Gordon Scharf, Jennifer YuI. Introduction:Sight and attention are valuable in the kitchen. Using tones to alert the cook is a common convention inappliance interfaces. A timer alert, however, is not a very informative form of feedback to receive from anappliance. Color is a better indicator of whether or not the food in question is ready. For the purposes of thisproject, we restricted the problem to only one class of baked goods: white bread toast. We then used awebcam to acquire images of the inside of the toaster and used image processing techniques to analyze thechange in color. The ultimate objective of the image processing is to detect when the bread has reached itsdesired level of brown and thus, presumably, its desired toastiness. However, this question can also be seenas determining the point in time when the toast has reached the optimum color change from its original toastcolor. Thus, instead of dealing with absolute color measures, the image processing could merely beconcerned with sensing relative color differences. By sonifying this change in color, we hope to help theuser be more productive and consistently achieve optimal brownness of their baked goods.II. Equipment:Logitech QuickCam for NotebooksLaptopToasterBread III. Image Acquisition:JPG images were taken at 30 seconds intervals using a Logitech QuickCam for Notebooks. Originally wetook a set of data manually holding the camera directly against the glass. This produced sharp images, andthe color differences were readily apparent. However, holding the camera against the glass of the toasteroven turned out to be too hot for the camera to withstand for long periods of time. Thus, we mounted thecamera by clipping it onto a piece of curved plastic that was taped to the handle of the toaster oven. Thedistance from the glass caused the quality of the images to decrease slightly. To further enhance the differences between the images, we removed the IR filter from the camera. Thisinvolved disassembling the camera and breaking off the small red filter that was in place. We then took newsets of images with the modified camera. Unfortunately, removing the IR filter seemed to wash out theimage with white light. Luckily, the datasets still seem to be usable provided that the right heuristic wasused to detect changes in the colors of the bread as it toasts.Another large problem is inherent to toaster ovens. On the “toast” setting of a toaster oven, the toaster ovenstarts off cold and with its heating coils turned off. As the toaster oven heats up, the heating coils begin toglow brighter and brighter red; as their brightness increases, the light inside the toaster oven increases whichdramatically alters the quality of the images. IV. Color Spaces:Although to the human eye the color differences between white bread and toast are readily apparent, it ismore difficult to sense these differences using a computer. One of the difficulties with detecting colorchanges as the bread toasts is that the bread will be a shade of brown at all stages of toasting and thus, undersome color schemes, very similar in appearance. There are a multitude of color models that exist for imageprocessing. The picture information produced by the camera is transmitted in terms of RGB values. However, in order to track the browning of the bread, we converted our photo information from the RGB tothe CIELAB color space.IV.1: RGB1/13/09 6:16 PMSonorific ToasterPage 2 of 10http://www.princeton.edu/~malia/report.htmIn an RGB color scheme, colors are decomposed into different amounts of red, green, and blue. This is anadditive color model; thus white is the additive combination of equal amounts of all three colors while blackis the absence of colored light. When an image is read into MATLAB, it is read using the RGB colorscheme. In MATLAB, images stored in this format are stored as a set of three arrays (one per color) eachof which is 240x320 (the total number of pixels in the image). In an RGB color scheme, all of the shades ofbrown look remarkably similar. Since what distinguishes the browns from one another is primarily theamount of white present in the image, a different color scheme is more suited to the problem at hand.IV.2: CIELABCIELAB is an opponent color system in which colors are decomposed into values that are represented by theL*, a*, and b* axes. Color opposition models supposedly most closely approximate a human interpretationof color. According to Adobe, “[c]olor opposition correlates with discoveries in the mid-1960s thatsomewhere between the optical nerve and the brain, retinal color stimuli are translated into distinctionsbetween light and dark, red and green, and blue and yellow.” Additionally, unlike other color schemes,CIELAB aspires to be perceptually uniform.In the CIELAB model, the L* axis is the central vertical axis in the space and represents lightness; its valuesrange from 0 to 100 where 0 represents black and 100 represents white. Both a* and b* are color axeswhose values range from negative to positive values. The a* axis represents the opposition between red andgreen, and the b* axis represents the opposition between yellow and blue. The origin, which is located atthe intersection of the three axes, is a neutral gray color. The diagram below illustrates the CIELAB colormodel.V. Final ImplementationV.1 Edge Detection: Jen YuOriginally we planned on using edge detection for two potential purposes. The first was to detect thelocation of the toast so that we could distinguish between toast areas and toaster areas. This would beuseful in the processing of the toast’s color information. The second was to track the change in volume ofthe toast. After we toasted our first piece of bread, we realized that, in addition to color, volume changessignificantly as bread toasts. Thus, if we could track the movement of the edge of a piece of toast, wewould have another indication of toastiness.Ultimately, edge detection was not incorporated into our final project. Edge detection of edges in the imageswas not difficult to do (using Canny edge detection algorithm), but figuring out the appropriate threshold touse and figuring out which edges corresponded to the


View Full Document

Princeton COS 436 - Sonorific Toaster: Image Processing

Documents in this Course
Load more
Download Sonorific Toaster: Image Processing
Our administrator received your request to download this document. We will send you the file to your email shortly.
Loading Unlocking...
Login

Join to view Sonorific Toaster: Image Processing and access 3M+ class-specific study document.

or
We will never post anything without your permission.
Don't have an account?
Sign Up

Join to view Sonorific Toaster: Image Processing 2 2 and access 3M+ class-specific study document.

or

By creating an account you agree to our Privacy Policy and Terms Of Use

Already a member?