Image Processing
Image processing is the art of transforming 2D-arrays of picture elements pixels. While the list of popular image-processing transformations is long, you can break down the majority of algorithms into three categories based on how the algorithms access pixels: sequential, random access, or sliding windows.
Algorithms in the sequential category process pixels independently and do not care about the spatial location of those pixels in the image. Example sequential algorithms include adding two images, thresholding images, and gamma correction. The idea behind gamma correction is to adjust the slope of the change in intensity from white to black by changing the value of a pixel x to some new value f(x) based on the function f(x)=x1/. Gamma correction can be used to increase or decrease the overall brightness of an image.
Random-access algorithms determine the next pixel to visit based on some computation performed at the current pixel. An example of such an algorithm is tracking eyes on a face from image frame to image frame.
Sliding window operations, the most common image-processing algorithms, are conceptually defined as two-dimensional matrices (kernels) that are passed over the image. The kernel is centered over each pixel in the image and some type of operation is performed over the pixels lying under the kernel. The result of the operation is a new pixel value. Examples of typical kernels are 3×3 or 5×5 squares, however, kernels are not restricted to fixed sizes and can be of any arbitrary shape. Example operations are convolution (where each pixel in the kernel is multiplied by the corresponding image pixel and the result is summed) and median filtering (where each pixel in the image from the kernel is sorted and the median is returned as the best value). Figure 2 shows an image with speckle noise and Figure 3 shows the result of a 3×3 median filter.
V.A., M.R.S., and J.B.P.