transform_deltaE.md

Calculate Delta E between observed and expected color cards

Calculates Delta E between a Macbeth ColorChecker or Astrobotany.com Calibration sticker style color card and the expected color values.

plantcv.transform.deltaE(rgb_img, color_chip_size=None, roi=None, obs="calibrated", method="deltaE_ciede2000", **kwargs)

returns Delta E matrix

• Parameters

  • rgb_img - Input RGB image data containing a color card.
  • color_chip_size - Type of color card to be detected, ("classic", "passport", "nano", "mini", "cameratrax", or "astro", by default None) or a tuple of the (width, height) dimensions of the color card chips in millimeters. If set then size scalings parameters pcv.params.unit, pcv.params.px_width, and pcv.params.px_height are automatically set, and utilized throughout linear and area type measurements stored to Outputs.
  • roi - Optional rectangular ROI as returned by pcv.roi.rectangle within which to look for the color card. (default = None)
  • obs - label for metadata and debug images, typically "calibrated" or "uncalibrated" depending on whether the rgb_img has been color corrected.
  • method - Function name from skimage.color as a string to use to calculate delta E. Currently deltaE_cie76, deltaE_ciede2000, deltaE_ciede94, and deltaE_cmc are supported
  • **kwargs - Other keyword arguments passed to cv2.adaptiveThreshold and cv2.circle.
    • adaptive_method - Adaptive threhold method. 0 (mean) or 1 (Gaussian) (default = 1).
    • block_size - Size of a pixel neighborhood that is used to calculate a threshold value (default = 51). We suggest using 127 if using adaptive_method=0.
    • radius - Radius of circle to make the color card labeled mask (default = 20).
    • min_size - Minimum chip size for filtering objects after edge detection (default = 1000)
    • aspect_ratio - Optional aspect ratio (width / height) below which objects will get removed. Orientation agnostic since automatically set to the reciprocal if <1 (default = 1.27)
    • solidity - Optional solidity (object area / convex hull area) filter (default = 0.8)

• Returns

  • deltaE - Delta E values per each color chip as a matrix.

• Context

  • Delta E is a perception-based metric for the difference between colors. 0 indicates no perceptual difference and higher values indicate more difference. Generally a delta E value less than 1 is imperceptibly different and values greater than 3.5 are clearly distinct. Whether a particular color chip's delta E value matters for your experimental goals depends on your hypothesis and analysis plan.

!!! note Delta E is calculated when a color card is detected with plantcv.transform.detect_color_card by default and a debug image is generated showing the differences in the color card against the expected colors. This function will use plantcv.transform.detect_color_card to find the color card, potentially in an image that has already been color-corrected so that the delta E values can be compared pre vs post calibration.

from plantcv import plantcv as pcv
rgb_img, path, filename = pcv.readimage("target_img.png")

pcv.params.debug = "plot"
# Delta E debug visualization shown below
cc_matrix = pcv.transform.detect_color_card(rgb_img=rgb_img)

# Next we may color correct the image
tgt_matrix = pcv.transform.std_color_matrix(pos=3)
corrected_img = pcv.transform.affine_color_correction(rgb_img=rgb_img,
                                                      source_matrix=cc_matrix,
                                                      target_matrix=tgt_matrix)

# Delta E on the corrected image
e_matrix = deltaE(corrected_img)
# outputs metadata will have min, mean, max, std dev of deltaE
print(pcv.outputs.metadata["max_deltaE_uncalibrated"])  # the uncalibrated metadata is added by detect_color_card
# [np.float64(49.5516548320252)]
print(pcv.outputs.metadata["max_deltaE_calibrated"]) # the calibrated metadata is added by deltaE
# [np.float64(15.730570870191682)]

Calling plantcv.transform.detect_color_card on the uncorrected image:

Using plantcv.transform.deltaE on the corrected image to check the new Delta E values: