Why is 450 nm monochromatic light perceived as blue or violet depending on its intensity?

This is a general feature of human perception of blue-white mixtures. It's known as Abney effect. It's not limited to highly monochromatic blue colors (not even to blue colors). We can observe this blue-related phenomenon even with longer wavelength light than 450 nm – e.g. sRGB blue, whose dominant wavelength is about 464 nm. Here's what most people would undoubtedly call blue – in HTML notation #0000ff:

And here's a mixture of 60%^† white with 40% blue (in HTML notation #cacaff):

To me, if I haven't stared at shades of truly violet a few moments ago, this mixture seems to have a considerable amount of violet — more than I'd expect looking at the primary blue. Naturally, shorter wavelength light should give even more violet perception, until it becomes violet even in its pure form.

---

^{† Percentage is given in linear scale, actual color presented is gamma-corrected to be displayed on sRGB monitors.}

This merely corroborates the experience but is in no way a bioengineering or biology explanation.

If you stare, at arm's length, at any discrete colour 5mm LED of high brightness, it's colour will appear to shift after a few minutes. This is not the (scotopic/photic) night vision effects from Cones to Rods towards blueshift but rather the cone's protective response to the high-density stationary colour indicators and makes that spot less-sensitive.

Also if one often stares at the bright ceiling or street lights for several minutes, then look away, they will see a remanent ghost image with that previously dominant spectrum now suppressed so it will appear in the inverse pastel colour.

It appears as the spectrum has been diminished in that spot, because it has, by a reduced sensitivity of the cones in the eye for detecting that spot of light. The recovery time is about the same time or a bit more as it takes to react. The human retina contains about 120 million rod cells and 6 million cones .

This may be safe as long as the light is not held too close to the eye and certainly not looking into a direct laser beam, rather a diffused image.

Even some 5mm LEDs at 20mA can be so intense as to be painful quickly at arm's length with >20,000 millicandella luminous intensity measure by std methods. But it only needs to be as bright as an auditorium ceiling light intensity for this example of color-shift to occur.

Added

Each depends on the spectrum of dominant and sub dominant peaks in the spectrum that results in the apparent shift in appearance. The 450nmD must have this sub dominant spectrum in an otherwise monochromatic spectrum for this affect to occur. Because our eyes are logarithmic yet spectral densities of monochromatic light sources are displayed on linear scale, it is often not reported. I.e. there must be some red 6xx nmD spectrum for this blue to appear to become purple when our eyes attenuate the blue after a period of intense viewing for several minutes.