java - Simple metric for blur in images? -

i'm trying tell if given photo blurry. know impossible task, , metric return undesirable results sometimes.

i'm wondering if there's simple metric @ least tries estimate blur can use though. specifically, task has high tolerance false positives. e.g. if got eliminated 90% of blurry photos , 50% of non-blurry photos happy.

i'm trying implement in java. have array of pixels (as ints). please keep in mind have limited understanding of image processing techniques (fourier transforms, etc.), , love specific walkthrough of how code solution.

as you've said you're not going find universal metric.

also there different types of blur: uniform, anisotropic, motion blur...

in general blurred images tend exhibit low frequencies. possible descriptor sum of magnitude of k highest frequencies. image low sum blurred overall.

the magnitudes can obtained in n*log(n) time using fourier spectrum (high frequencies far origin) or laplace pyramid (high frequencies correspond first scales).

wavelet transform possible descriptor