Agricultural crop cultivation requires proper care and maintenance to minimize the risk of plant diseases, which can significantly impact crop yields. This study aims to compare two algorithms to identify the most effective approach for plant leaf disease classification. The research utilizes a curated dataset consisting of images of six different plant leaf diseases, totaling 21,733 images. The algorithms employed in this study are computer vision-based techniques. The first method involves image preprocessing, including resizing, brightness and sharpness adjustments, and color transformation using the CIELAB color space. Additionally, K-means clustering is applied to enhance key visual features in the images. The second method focuses on anomaly detection, where Mahalanobis distance weighting is used to highlight anomalous regions within the images. Once the images are processed, they are fed into a machine learning model for classification. This study employs the Vision Transformer (ViT), which is capable of effectively detecting and classifying plant leaf diseases due to its ability to capture long-range dependencies within images. This show effectively highlights distinct regions within the image prior to classification, outperforming the LabK-means at 93% compared to Mahalanobis Distance at 76.40%.