Classification of citrus leaf diseases by image processing based on texture (Statistical) related features

Abstract

The whole image has plenty amount of pixels, but only a small amount of pixels are essential and those only required to be separate out. In image processing, feature extraction is a special technique for dimension reduction. Transforming the input data into a small and meaningful set of features is called feature extraction. If the features extraction process is carefully chosen then it is expected that the feature set will find the relevant information from the input image in order to perform the desired task using the reduced formation of features instead of the full size input image. In machine learning, pattern recognition and in image processing, feature extraction starts from an initial set of measured data and builds derived values (features) intended to be informative and non-redundant, facilitating the subsequent learning and generalization steps, and in some cases leading to better human interpretations. As a human being one can better understand the reduced feature set instead of the whole image. The paper presents total thirteen texture related features like: Contrast, Correlation, Energy, Homogeneity, Mean, Standard_Deviation, Entropy, RMS, Variance, Smoothness, Kurtosis and Skewness. These texture features extracted by using four leaf disease segmentation techniques and they are: Segmentation with RGB and HSI, K--Means Clustering Algorithm, Segmentation by Different Transformation and Segmentation by RGB, HIS and Contrast respectively. Finally the extracted features applied to neural network with three training functions and results analyzed. The paper compares the benefits and limitations of these potential methods.