A Novel Approach for Pomegranate Image Preprocessing Using Wavelet Denoising

Abstract

Image preprocessing is the first and foremost unit in any image processing/machine vision application, because preprocessing inspirits the accuracy of the other imaging processes such as morphological processing, segmentation, feature extraction. The major difficulty in preprocessing is the variation in the levels of intensity values of the image and the presence of noise. Therefore, histogram equalization needs to be performed prior to any further processing of the image. This paper reports on the development of a novel approach for enhancement of the pomegranate images. The major contribution of the present paper is to perform histogram equalization of pomegranate images in combination with wavelet denoising. The proposed method not only enhances contrast through histogram equalization but also has advantage of compensating for the loss of information of the images. The paper compares experimentally the informational entropy of the images. Different alternatives of preprocessing were applied, and informational entropy was computed. To evaluate the robustness and accuracy of the proposed method, tests were conducted for 166 sample images. The results showed an accuracy of 90% success rate in retaining informational entropy of the images when wavelet processing was applied in combination with histogram equalization. The results illustrate that the proposed method can be easily extended to other image processing-based agricultural applications.