Background

Accurate quantification of plant disease severity is fundamental for studying plant–pathogen interactions, evaluating disease management strategies, and screening host resistance. Traditionally, disease severity has been assessed using visual rating scales or by estimating the percentage of symptomatic tissue on plant organs. Although such approaches are widely used in plant pathology, they are inherently subjective and may introduce observer bias and variability among evaluators [1,2]. These limitations can reduce reproducibility and may affect the statistical interpretation of disease assessment in experimental studies.

To overcome these constraints, digital image analysis has increasingly been adopted as a more objective and reproducible approach for plant disease phenotyping. Image-based disease assessment enables automated detection and quantification of symptomatic regions by analyzing color, texture, and spatial characteristics of infected tissues. Several studies have demonstrated that image analysis methods can significantly improve accuracy and consistency in disease measurement compared to visual estimates [3,4]. Such approaches are particularly effective for diseases that produce visible symptoms such as necrotic lesions, blights, spots, or rotting tissues that contrast clearly with healthy plant tissue.

Among the different approaches used in image-based disease detection, color-space segmentation has proven to be particularly effective for separating symptomatic and healthy regions in plant images. The hue-saturation-value (HSV) color space is commonly used in plant image analysis because it separates chromatic information from intensity, making it more robust to variations in illumination compared to the traditional RGB color model [5]. HSV-based segmentation has been successfully applied in several plant phenotyping studies, including leaf segmentation, stress detection, and disease symptom quantification.

Image analysis pipelines often combine color segmentation with morphological operations to improve the accuracy of tissue classification. Morphological image processing techniques such as opening, closing, and hole filling are widely used to remove noise, refine segmented regions, and generate continuous masks representing diseased areas [6]. When combined with pixel-based area calculation, these techniques allow precise measurement of diseased tissue relative to total leaf area, enabling quantitative estimation of disease severity.

Various image analysis platforms, such as ImageJ, specialized plant phenotyping software, and machine learning–based tools have been developed for plant disease quantification [4,7]. However, many of these systems require specialized plugins, large training datasets, or complex workflows that may limit their accessibility for routine experimental use. In contrast, simple threshold-based segmentation approaches implemented in scripting environments such as MATLAB provide a flexible and reproducible alternative for disease quantification when symptoms exhibit clear visual contrast with healthy tissue and images are acquired under reasonably controlled conditions. The present protocol does not introduce a new image-processing algorithm; rather, it provides a detailed, step-by-step implementation of an established HSV threshold-based workflow that can be readily adopted, reproduced, and adapted by researchers for routine disease quantification.

The protocol described here provides a MATLAB-based workflow for quantitative differentiation of diseased and healthy plant tissue from digital leaf images using HSV color segmentation combined with morphological image processing. The method involves the acquisition of standardized leaf images, the conversion of RGB images to HSV color space, the segmentation of diseased regions based on defined threshold ranges, the refinement of the segmented mask using morphological operations, and pixel-based quantification of diseased and total leaf area. The protocol requires MATLAB together with the Image Processing Toolbox and does not depend on machine learning models or external image-analysis software.

Although the method was developed using citrus canker symptoms, the workflow may be adapted to other plant diseases that produce visually distinguishable symptoms, such as necrotic lesions, blights, leaf spots, anthracnose, or other forms of conspicuous tissue discoloration. Because segmentation relies on fixed HSV threshold values, users may need to recalibrate the threshold ranges and associated image-processing parameters when applying the protocol to different plant species, disease symptoms, imaging systems, or illumination conditions. Accordingly, this protocol is intended primarily as a practical, reproducible workflow for disease quantification under reasonably controlled imaging conditions rather than as a universally applicable segmentation method.