means and standard deviations of hsv histograms are not typically all that meaningful. What question would such a measurement be trying to answer??
Suppose you take
data = rand(1000,1000,3);
rgb = hsv2rgb(data);
That is, we generate uniform data in hsv space, such as would be needed for it to be reasonable that a histogram of the v component would be more or less equally distributed randomly. We convert the uniform data into hsv space, and then look at the red, green, and blue histograms of the rgb data. We can see that most of the data is towards the left edge for all three components.
Thus, in order to expect uniform v component you need to be starting with rgb that tends towards the darker.
If you take that same data and do
and plot the third component, you can see that it is biased towards high value, so rgb that is uniformly random is mostly towards the high value. (Uniform random rgb data is bell curve in brightness, by the way.)
I ran through all of the color demonstration images provided with the image processing toolbox. The one with the smallest standard deviation was onion.png -- which has a small deviation mostly because it is one of the smallest of the images. eSFRTestImage.jpg has the largest standard deviation -- mostly because it is one of the largest images.