The objective of this research was to develop an image analysis technique for severity rating of
brown leaf spot disease in cassava. Samples of cassava leaves were collected from field and imaged
under controlled illumination. Images resolution were resized to 640×480 pixels and transformed from
RGB to HSI color space. The transformed images were then segmented and feature-extracted in order
to determine total leaf area and diseased area. Noise reduction was performed using erosion and
dilation procedure. The percentage of diseased area was calculated based on diseased area and total
leaf area. Comparison of number of spot counts with visual observation indicated an agreement at an R2
= 0.90, besides, the number of spots was found to affect the accuracy of image analysis. Percertage of
disease area was consistent with grid counting method, giving an R2 = 0.78.