Digital image processing involves sampling images by representing each pixel as a value and quantizing the number of photons hitting each sensor to a number. Sampling can cause aliasing and downsampling reduces image size and information, while upsampling increases size but not information. Images are stored as matrices of values that may include multiple color channels represented by bits per channel. Requantization changes these representation details like converting color to grayscale.
Digital image processing involves sampling images by representing each pixel as a value and quantizing the number of photons hitting each sensor to a number. Sampling can cause aliasing and downsampling reduces image size and information, while upsampling increases size but not information. Images are stored as matrices of values that may include multiple color channels represented by bits per channel. Requantization changes these representation details like converting color to grayscale.
Digital image processing involves sampling images by representing each pixel as a value and quantizing the number of photons hitting each sensor to a number. Sampling can cause aliasing and downsampling reduces image size and information, while upsampling increases size but not information. Images are stored as matrices of values that may include multiple color channels represented by bits per channel. Requantization changes these representation details like converting color to grayscale.
Digital image processing involves sampling images by representing each pixel as a value and quantizing the number of photons hitting each sensor to a number. Sampling can cause aliasing and downsampling reduces image size and information, while upsampling increases size but not information. Images are stored as matrices of values that may include multiple color channels represented by bits per channel. Requantization changes these representation details like converting color to grayscale.
◦ Computer vision as preprocessing step ◦ Each sample is represented as a pixel
◦ Images compression ◦ In acquisition devices, the image sensors contain sepa- rate elements for each pixel ◦ Has many active research areas ◦ During sampling, each element will record the amount of photons hitting the area ◦ As computer vision gets more important, so does image processing ◦ During sampling, each element will take photons from multiple objects, if they are close to each other. This is the reason why acquired images do not have extremely Image sharp edges ◦ Changing the sampling rate of the image is called re- ◦ Abstract or realistic representation of a scene sampling
◦ It could be a picture or a digital image ◦ Up sampling (zooming) can be used to increase the size of the image, however, this will not increase the amount ◦ There are two types of digital images of information contained within the image Continuous / parametric: called a vector image ◦ Down sampling (shrinking) can be used to decrease the size of the image, and generally will result in the loss of Discreet Images: Fixed sized and represented as information matrices, we will only work on discreet digital im- ages ◦ Sampling can cause strange spatial effects like aliasing and moire effect
Discreet images Quantization ◦ Entire image is stored as a matrix ◦ The operation of converting number of photons hitting ◦ Matrix cells can contain a single value or a group of a sensor element to a number values for multiple color channels ◦ Generally, images are saved as 8-bit RGB (red, green, blue) format which requires 3 bytes per sample ◦ Generated (CGI, computer generated image)
Rasterized from vector images
◦ Each different color value stored per pixel is called a channel Generated from fractals or other mathematical for- mulas ◦ Average, minimum and maximum intensity should be considered while performing quantization A rendered 3D scene Common properties include sharp edges and flat ◦ In an 8-bit image, maximum number that a pixel can get is 255. This represent the maximum intensity colors Too clean ◦ In a black and white image, maximum intensity signifies white color, whereas 0 represents black ◦ Acquired images ◦ Quantization can cause over and under exposed images Using a camera or a scanner as well as inaccuracies under low light conditions
Has softer features
Contains naturally occurring noise as well as noise Requantization caused by acquisition process ◦ Changing bits/channel or channels of a digital image Acquisition process reduces the amount of details in the real scene ◦ The transformation can be performed to change color Fully capturing a real 1 m2 scene would need 1071 image to grayscale or vice versa pixels ◦ Different color representations can be converted to each Thus the amount of details would be lowered and other this process is called sampling ◦ Converting a color or grayscale image to a mono-chrome Restricting the acquired brightness level to a finite (black and white image) is also a requantization opera- range is called quantization tion
St. Joseph'S College of Engineering, Chennai - 119. Department of Science Ph8252 - Physics For Information Science Worksheets Unit - Iv Optical Properties of Materials
