Discrete Cosine Transform

Abstract

Moving continuous tone images are represented as a sequence of “frames”. A frame is a two-dimensional array of pixel values in one “plane” for black and white images, or more planes for color images. We model the signal being sampled (a sequence of pixel values forming a row, column, or time-varying sequence) as a random variable with a mean of zero. The probability distribution, shown in Figure 5.1, of pixel x₁ given the value of pixel x₀ has been shown empirically to be an exponential (laplacian) distribution [RY90]:

$$ P(X = {\mathbf{ }}\chi _1 {\mathbf{ }} - {\mathbf{ }}\chi _0 ) = {\mathbf{ }}\frac{{e^{ - \lambda |\chi _1 {\mathbf{ }} - {\mathbf{ }}\chi _0 |} }} {{2\lambda }} $$

((5.1))