Preprocessing of tomato images captured by smartphone cameras using color correction and V-channel Otsu segmentation for tomato maturity clustering

Abstract

Preprocessing stage is an essential part in image processing or image recognition. Image taken by smartphone cameras may have inconsistent color that leads to inconsistent intensities, although they are captured in the same position and lighting condition. Apart from color inconsistency, there is a probability that smartphone camera produces blurry images. In order to solve those problems, this paper proposes a new framework to preprocessing image using combination of Linear Regression algorithm and V-Channel Otsu segmentation. Color correction and V-Otsu segmentation yield better segmentation and achieve good results after being evaluated using 6-means clustering. There are four types of smartphone devices tested to capture all tomato images. Since not all devices produce clear images, to test blurred image we use the variance of Laplacian. Based on experiment, Samsung Galaxy Ace produces the most blurred images. Preprocessing applied in blurred images using combination of Linear Regression and V-Channel Otsu segmentation (LR-V-Otsu) yield MSE up to 1.033. This result concludes that the algorithm is robust for blurred image.