Extracting Color Features of Leaf Color Images

In this paper, several methods that were not influenced by shapes, sizes of leaves and image background noise were applied in extracting color features of tomato leaves that were short of a kind of nutrient. The methods included statistical algorithms, a algorithm of the correlation coefficient between two different color matrixes of the same leaf and the available ways of extracting color features by their histogram shapes. But these methods were not effective in extracting the color features of the tomato leaves. A new way of percent histogram was put forward. It includes the research of getting rid of the disturbance of the white background noise in the leaf image by drawing the histogram, and the research of removing the interference of leaf size by changing the histogram into percent histogram and the way of looking for the better color position in percent histogram for extracting effective color features. The experimental results of extracting color feature of the tomato leaves proved that the method was effective in separating different kinds of tomato leaves, and showed that the accuracy was above 70%.