Measuring area of colour patches (e.g. dead or damaged areas) on leaf surfaces



Kim Pullen, Saul Cunningham


This protocol was used to determine the area of necrotic tissue on a large sample of leaves. The same approach could easily be adapted to measuring the areas of other distinct patches defined by colour variation.


Many eucalyptus species suffer high levels of insect damage to leaves. In our study we wanted to quantify whether an insecticidal treatment reduced insect damage on Eucalyptus blakelyi. We observed that insect damage could reduce the number or size of leaves, and could also reduce the effective area of the remaining leaves by causing leaf necrosis. In our case the necrotic tissue was characterized by brown, tan and red colouration (Fig 1), but from a leaf function point of view the key observation was that it was not green, whereas undamaged leaf blade was green.

We determined that, for example, unprotected leaves on the south side of the tree canopies had a median level of approximately 25% leaf necrosis by area (Cunningham et al. 2009).

 - File

Fig 1: Some leaves can have large areas of necrosis due to insect damage. In this case (Eucalyptus blakelyi) the necrotic areas shows a range of colours, but clearly is deficient in green.


  • Flatbed scanner (e.g. Hewlett Packard Scanjet 2400)
  • ImageJ image analysis software
  • A spreadsheet package (e.g. Excel)

Units, terms, definitions




  1. Scan fresh leaves on a flatbed colour scanner, acquiring colour images. Save in jpeg format. Make sure to use a contrasting background colour. Either scan one leaf at a time, or scan multiple leaves but analyse each leaf separately.
  2. Print a sample of your leaf images in colour onto a sheet of plain copy paper.
  3. For each image, cut the image out with a pair of scissors to remove the background. Cut out green and necrotic areas (judged by eye) and separate the bits of paper into two piles. Keep each leaf separate.
  4. Separately weigh the “green” and “necrotic” bits of paper. Use the weight as a relative estimate of area. Record the weights and use these to calculate the % necrotic area for your test sample leaves.
  5. Load an image (from step 1 above) into ImageJ by selecting open from the file menu (or Ctrl+O). Select the area of the image that you want to analyse (e.g. the whole leaf, or part of the leaf surface) using the click and drag selecting rectangle, which is open by default. Alternatively, you may be able to use the magic wand tool to select the leaf only.
  6. Then select the Histogram command from the Analyze menu (or Ctrl+H) to open the histogram window. From this window you can press the “copy” button, and then paste the data into your spreadsheet, or you can press the “list” button and then save the data as a file.
  7. The output has two columns; the first column is the level (255 levels of colour in this case) and the second column is the count of pixels at that level (we refer to this as the pixel table). Images saved at higher resolution will have more pixels, but the relative frequency at each level should be similar to that seen at lower resolution.
  8. Select an area of leaf that you would like to classify as “good’, and generate the pixel table.
  9. Select an area of leaf that you would like to classify as “necrotic’, and generate the pixel table.
  10. Repeat for a sample of good and necrotic patches for each of the leaves in your test sample, keeping note of which lines of data belong to which selections.
  11. Using a spreadsheet (we used excel) visualize the data by generating separate histograms for the “green” and “necrotic” areas. Although the overall visual impression will be that they are distinct, it is likely that some colours occur in both good and necrotic areas creating overlap in the pixel count histograms of green and necrotic patches. At this stage we summed across a set of green patches and across a set of necrotic patches to create single aggregated histograms based on multiple samples. By inspecting your data, decide on a few different criteria that could be used to assign colour levels into categories (i.e does that colour level count as green or nectrotic ). Because of the overlap described above, this step requires approximation and judgment. In our study the categories were: 1) green leaf surface, 2) not-green leaf surface (i.e. necrotic) and 3) background (i.e. not leaf). Use what you see in your histograms as the basis for testing different categorization approaches (see next step).
  12. One at a time, analyse the whole leaf image for your sample of test leaves and save the pixel table for each one.
  13. In your spreadsheet sum the number of pixels in your categories (green, not-green leaf, background) and calculate the percentage necrotic area per leaf. Repeat this for your whole sample of test leaves using a number of different criteria for categorization.
  14. Examine the correlation between the % necrotic area per leaf using the “by eye” method to the % necrotic area using the pixel count method. Test this correlation for each of the different criteria for categorization that you are considering. This will tell you first whether there is a good fit between the pixel count approach and the by eye approach, and second will help you chose which criteria for categorization give you the best fit of all. In our case, for example, we categorized green tissue as pixels with the colours 75-140, white background was 242-255, and everything else contributed to the necrotic category. Using these criteria the correlation with the by eye method was r = 0.95.
  15. By this stage you know how to capture images, generate a pixel count, and convert the pixel count into an estimate of green or necrotic tissue. Now you can tackle the real sample scans of all your leaves, and start adding up the numbers.

ImageJ image analysis software (free download)(external link)

Literature references

Cunningham SA, Pullen K, Colloff MJ 2009 Whole-tree sap flow is substantially diminished by leaf herbivory Oecologia 158: 633-640 (This paper uses the above protocol)

Leave a Reply