It is well known that the human visual system is quite sensitive to color. The number of distinguishable intensities, for example, is much smaller than the number of distinguishable colors and intensities. In addition, color images are generally much more pleasant to view than black-and-white images. The aesthetic aspect of color can be used for image enhancement. In some applications, such as television commercials, false color can be used to emphasize a particular object in an image. For example, a red banana in a surrounding of other fruits of natural color will receive more of a viewer’s attention. In other applications, data that do not represent an image in the conventional sense can be represented by a color image. In this case, the color used is called pseudocolor. As an example, a speech spectrogram showing speech energy as a function of time and frequency can be represented by a color image, with silence, voiced segments, and unvoiced segments distinguished by different colors and energy represented by color brightness.
The use of color in image enhancement is limited only by artistic imaginations, and there are no simple guidelines or rules to follow. In this section, therefore, we will concentrate on three examples that illustrate the type of image enhancement that can be achieved by using color, In the first example, we transform a monochrome (black and white) image to a color image by using a very simple rule. To obtain a color image from a monochrome image, the monochrome image is first filtered by a lowpass filter, a bandpass filter, and a highpass filter. The lowpass filtered image is considered to be the blue component of the resulting color image. The bandpass filtered image is considered the green component, and the highpass filtered image is considered the red component. The three components-red, green, and blue-are combined to form a color image. Figure 3.49(a) shows an original monochrome image of 512 x 512 pixels. Figure 3.49(b) shows the color image obtained by using this procedure. The color is pleasant, but this arbitrary procedure does not generate a natural-looking color image. Changing classic black-and-white movies such as Casablanca or It’s A Wonderful Life to color movies requires much more sophisticated processing and a great deal of human intervention.
In the second example. we consider the display of a 2-D spectral estimate on a CRT. The 2-D spectral estimate, represented by Px(w1,w2) in dB, is typically displayed by using a contour plot. An example of a 2-D maximum likelihood spectral estimate for the data of two sinusoids in white noise is shown in Figure 3.50(a). The maximum corresponds to 0 dB and the contours are in increments of 0.5 dB downward from the maximum point. In such applications as detection of low-flying aircraft by an array of microphone sensors, we wish to determine the number of sinusoids present and their frequencies. An alternative way of representing the spectral estimate is to use pseudocolor. Figure 3.50(b) gives an example, where different amplitudes of Px(w1,w2) have been mapped to different colors. Comparing the two figures shows that the two peaks and their locations in the spectral estimate stand out more clearly in Figure 3.50(b).