Idrisi Lesson 6  -  Creating Composite Images

Image enhancement is one of the four basic operations in digital image processing.  The other 3 operations include: (1)  restoration, what Campbell calls preprocessing - geo-referencing, removal of striping, and unwanted atmospheric effects; (2) transformation, in which the brightness values from various bands are mathematically combined to synthesize a new set of brightness values; and (3) classification, in which groups pixels of similar brightness characteristics together to aid in interpretation.

The goal of image enhancement is to alter an image to highlight visual differences.  Our eyes are not great at detecting variations in gray tones.   Once colors are added, our eye easily identifies patterns.  Image enhancement also helps in digital classification routines by altering brightness values to emphasize certain brightness ranges.  Enhancement processes includes contrast stretching, composite generation, and digital filtering.  Contrast stretching was a subject of an earlier IDRISI exercise.  Digital filtering is not covered in this first course, but includes, among other topics, "sharpening" the appearance of an image, and shading an image (directional filtering) to provide a sense of topographic relief.

Composite generation adds colors to the different bands to enhance differences in brightness values.  One way to create a true RGB (red, green, blue) composite, would be to sandwich, digitally speaking,   3 layers with each layer devoted to a single color.  For example, the red (or green, or blue) layer would be clear for no brightness (DN=0), and saturated red (or green, or blue, respectively ) for a brightness maximum (DN = 255). 

The resulting color of the image would be a combination of RGB colors.  We are limiting ourselves to 3 colors and therefore 3 bands.

Although we think of the red layer being red visible light (0.6-0.7 um), the green layer being green light (0.5-0.6 um), and so on, there is no reason we can put any band we want in the red layer, in the green layer, or in the blue layer.  As a matter of convention the shorter wavelengths are included in the blue layer and the longer wavelengths in the red layer.

Rather than creating a true RGB composite, IDRISI uses a short cut.   First, it stretches (compresses, actually) the brightness in each band from 256 (0-255) levels into 6 levels (0-5).  For example, if the brightness were a maximum for each layer (red = 255; green = 255; blue = 255), the stretched result for each layer would be red = 5, green = 5, and blue = 5. 

Then it mathematically combines the 3 bands to form a new set of brightness values:

new brightness = blue + (green*6) + (red*36)

Continuing with our example, we have,

new brightness  =   5 + (5*6) + (5*36)

                        =  215      

Another example, red = 40; green = 255; blue = 170), the stretched result for each layer would be red = 1, green = 5, and blue = 3. 

new brightness  =   3 + (5*6) + (1*36)

                          =  69

Although this approach is, technically speaking, image transformation, the intent of the method is to mimic a true RGB color composite and is therefore included within the enhancement subcategory of  digital processing.

Another reason for this approach is that IDRISI cannot perform unsupervised classification on a set of individual bands.  The original set of bands must be reduced and combined into a single composite.

The final step is to assign a color palette to the new set of brightness values.  The particular combination of the stretching, mathematical combination of values, and color palette used by IDRISI is designed to mimic a true color composit as described above.

Question:

1. (5 pts) In the stretched version of each band, BEFORE they are combined, is information gained, lost, or unchanged.  Explain.

STEPS FOR CREATING A COMPOSITE IMAGE:

1. Command:   ANALYSIS – IMAGE PROC — ENHANCEMENT — COMPOSIT

2. Composite images for use in unsupervised classification are constructed using the three most “informative" bands. Depending on the composite, one always wants to use the most informative bands.  Look back to your PCA results from Exercise 5.  You will note that the most informative bands are 2, 3, and 4.  Use bands xmex-b2, xmex-b3, xmex-b4.

Question: 

2. ( 5pts) Why are not both bands 1 and 2 used?

3. For an output image name, use "mexcmp6".

4. Select linear w/ saturation.  Specify 1% (1% is standard for use with unsupervised classification).

5. Enter a title for the image. 

6. Select OK to process.

Display composite images using the color composite 256 palette. Thinking back to the gray tone color scales of the previous exercise, the color version here certainly highlights the difference between hillsides and valley bottoms. Other band combinations, can also be used to produce other valuable results.

More Questions:

3.      (1 pts) What is the color of the valley bottom __________________________.

4.      (1 pt)  What is the color of the hillsides?  _____________________________.

5.      (2pts)  Thinking back to Dr. Hadley's slide presentation, what is the difference in vegetation between the valley bottoms and the hillsides?

6.      (2 pts)  What is this false color image of this part of Mexico, highlighting?