Phenocam color image calibration using image analysis

Abstract.

Phenocam is a network of time-lapse digital cameras installed all over North America producing images to monitor the phenological changes in plants and correlating it to the climate change patterns. Color is an important discerning feature in an image. The application of image processing in agriculture mostly relies on correlating color changes to plant health, crop stress, or nutrient deficiency. Changes in lighting conditions during image acquisition affects the color attribute even though there is no change in the plant quality. Hence, there exists a need to calibrate images to bring them to common ground for better phenological comparison. Thus, in the present study, the focus is on developing a methodology to calibrate the image color values so that the images look balanced, irrespective of the input lighting conditions. For calibration, a standard color panel with 24 patches of known color values was included in all the images, obtained using a digital camera for different image sources and lighting conditions. A user-coded ImageJ plugin was developed in Fiji for image calibration. The plugin developed a relationship between the color values of each color patch in the input image to the standard color values from x-rite ColorChecker and a 3x3 color calibration matrix (CCM) was derived. The CCM was then applied to all pixel values in the input image, thus producing a calibrated image. The developed plugin performed well and took approximately 15 s to produce the calibrated image. New performance parameters such as difference index, and calibration performance index (CPI) were developed to evaluate the calibration performance. Best calibration was characterized by minimum difference index, and CPI value. The minimum number of color patches required for efficient color calibration was also determined by including color patches one by one in different orders. The results revealed that, usage of three additive primary color patches (red, green, and blue) are sufficient for a good color calibration.