Digital Image Processing Lecture 26: Color Processing Prof. Charlene Tsai
Color Model (Review) We’ll focus on RGB and HSI models in this lecture RGB: Red-Gree-Blue model for color monitors and color video cameras HIS: Hue-Intensity-Saturation model. Color and gray-scale information decoupled, so suitable for many existing gray-scale techniques.
RGB Model HIS Model
Exercise (RGB->HIS)
Color Transformations Techniques that process the components of a color image within the context of a single color model, as apposed to conversion between models. Techniques of interest Color complements Color slicing Histogram processing
Color Complements Analogous to gray-scale negatives Similar to conventional color film negatives Directly opposite of another on the color circle
Color Slicing Highlighting a specific range of colors in an image Separating object from their surroundings The simplest is to define the range of interest by a cube, or a sphere for sphere
Color Slicing - Example sphere cubic
Histogram Equalization Review: producing an image with an uniform histogram of intensity values. How to go about doing it? Erroneous if performing HE on individual color component. More logical in HIS space Hue and saturation unchanges HE on color intensity
Histogram Equalization - Example Before HE, median=0.36 After HE, median=0.5
Smoothing – Neighborhood Averaging RGB: each component can be smoothed independently HIS: smoothing only the intensity component (so more efficient)
Sharpening – Laplacian Enhancement RGB: computing the Laplacian of each component independently HIS: Computing only the Laplacian of the intensity component
Color Segmentation in RGB Works better than HIS model more systematic, less application-dependent Given a set of sample colors of interest: compute the average vector a for each pixel, determine if the color is in specified vicinity D0 of a the similarity measure is the Euclidean distance Very similar to color slicing
(con’t) C is the covariance matrix of the samples. If C=I , D(z,a) is reduced to |z-a| (the Euclidean distance). Reducing computation
Color Edge Detection There are many ways of doing edge detection on color images Method I: generating the gradient information on individual planes and combining the results Method II: computing the gradient of vector c at any point (x,y)
Color Edge Detection: Method II Let r, g, and b be the unit vector along the R,G, and B axis of the RGB color space Define the vectors: Let the quantities gxx, gyy, and gxy be
(con’t) The direction and magnitude of max rate of change of c(x,y) is
Result of using Method I
Method II Diff. Method I