Color rotation in HSV using scikit-image

The aim is to convert a pure red image into any hue of the color wheel.

A monochrome image is first converted into a RGB red image, ex:
then converted into HSV
The hue component is modified by adding an angle value supposed to match the wheel color
then the hsv image is back converted into the rgb color space.

The problem is that only green or blue image can be obtained (no yellow for an angle~30° for example):

The code performed in some ipython cells relies on scikit-image 0.10dev:

from skimage import io
from skimage import color
from scipy import ndimage as nd
import numpy as np
from matplotlib import pyplot as plt
import os
cy55 = io.imread('/home/jeanpat/MFISH/PSI/P07/01/Cy5/P070109C.tif')

zero = np.zeros(cy55.shape,dtype=np.uint8)
rgb0 = np.dstack([cy55, zero,zero])

hue_rotations = [18, 36,72,90,108]
images = {}
images[0] = rgb0
hsv0 = color.rgb2hsv(rgb0)
print hsv0[:,:,0].dtype
for hue in hue_rotations:
    hsv = np.copy(hsv0)
    hsv[:,:,0] = hsv[:,:,0]+ hue
    rgb = color.hsv2rgb(hsv)
    images[hue] = rgb
i = 1
plt.figure(num=None, figsize=(15, 5),  facecolor='w', edgecolor='k')
for hue in np.sort(images.keys()):
    plt.subplot(1,6,i,xticks=[],yticks=[])
    plt.title('hue='+str(hue))
    plt.imshow(images[hue])
    i = i +1
plt.show()

I answered this on the mailing list, but I'll copy the solution here so that it's easier to find (and the formatting is prettier).

Basically, there are some differences in how hue is represented (0--1 instead of 0--180), uint8 vs float data-types, and maybe some issues with how the grayscale image is converted to RGB. A quick example of usage might look like:

import numpy as np
import matplotlib.pyplot as plt
from skimage import color
from skimage import data


def colorize(image, hue):
    """Return image tinted by the given hue based on a grayscale image."""
    hsv = color.rgb2hsv(color.gray2rgb(image))
    hsv[:, :, 0] = hue
    hsv[:, :, 1] = 1  # Turn up the saturation; we want the color to pop!
    return color.hsv2rgb(hsv)


image = data.camera()[::2, ::2]

hue_rotations = np.linspace(0, 1, 6)  # 0--1 is equivalent to 0--180
colorful_images = [colorize(image, hue) for hue in hue_rotations]

fig, axes = plt.subplots(nrows=2, ncols=3)

for ax, array in zip(axes.flat, colorful_images):
    ax.imshow(array, vmin=0, vmax=1)
    ax.set_axis_off()

plt.show()

which gives: