Quantitative color image analysis of agronomic images

Many kinds of information that are available to our eyes in a qualitative sense require large investments of time and instrumentation to confirm quantitatively. If this information is accessible to our eyes, however, it should be possible to capture it in a color photograph and analyze it quantitatively. Our objective was to develop and make available software for enhancing and extracting quantitative information from color images. The resulting programs allow researchers to readily obtain quantitative data from color images such as photographs of plant canopies or soil with dye. The programs run in DOS and require only 512 kB of RAM, but can handle images up to 16 380 pixels across and 65 535 pixels down. The programs are designed to assist in developing an equation relating an independent parameter of interest (such as N content) to color parameters in calibration images, and then in using these equations to predict parameter values in new images. This allows one to convert, for example, an aerial photograph of a diseased wheat (Triticum sp.) field to a digital image in which the value of each pixel represents an estimated damage level. In a simple example use of the program, we show how percent canopy cover can be measured from color photographs. The programs, manual, and portions of the source code are in the public domain and are available at http://www.agron.iastate.edu/soilphsics/ewing_color.htmI.