Implementation of Histogram Equalization for Image Enhancement in The Classification of Spices Using K-Nearest Neighbor

Abstract

Purpose: To determine the effect of implementing Histogram equalization (HE) at the image preprocessing stage to improve image quality in rhizome spice classification using the K-Nearest Neighbor classification method.

Design/Method/Approach: Rhizome spice data was taken directly using a camera with a total of 600 images divided by a ratio of 80:20 for training and testing data. Preprocessing is done starting from resize to 512x512 pixels, then remove background to remove background objects that are not needed, then histogram equalization and also grayscale conversion. Glcm texture feature modeling, rgb color feature and hsv color feature are used as classification parameters. Classification is done using the K-Nearest Neighbor (KNN) method.

Findings/result: The test results of this study can be concluded that the application of HE at the image preprocessing stage succeeded in improving classification performance as seen from the accuracy evaluation value. In KNN classification without preprocessing histogram equalization gets an accuracy of 73.8%.  When implementing histogram equalization the classification accuracy increases to 76.1%.From the two accuracy results obtained, it can be seen that the implementation of histogram equalization has a good effect in increasing the accuracy of classification.

Originality/value/state of the art: The application of Histogram equalization (HE) in image preprocessing is able to improve image quality so that classification accuracy can increase compared to without using histogram equalization preprocessing.