Mask for watershed segmentation

[JorgeGtz]

The watershed segmentation as I've been using it requires the user to load all the images and stack them in a numpy array (grayscale). It also requires a stack of masks that separate the plants from the background.
Those arrays can be created with a loop that loads each image, converts it to grayscale, and also calls a function that performs color segmentation for the mask.

image_series = np.zeros((h,w,n_images))
mask_series = np.zeros((h,w,n_images))


# Load each image, transform it in gray scale and mask and stack them in their respective array
for i,image_name in enumerate(file_names):
    img, _, _ = pcv.readimage(filename=args.indir+image_name)
    img_gray = pcv.rgb2gray(rgb_img=img)
    image_series[:,:,i] = img_gray
    mask = create_mask(img)
    mask_series[:,:,i] = mask

I am considering changing the watershed function so it takes the path to the images instead of the array as argument. That way there would be less work for the user and the function could be used in longer time-series (by keeping all images in memory at once).

The problem is that the color segmentation for the mask still requires some tuning by the user.

[JorgeGtz]

The masks for the image series can be created in a PlantCV workflow and stored to then be used by a second workflow that uses the watershed segmentation.