def get_roi_im(w, h):
sz = FastImage.Size(w, h)
roi_im = FastImage.FastImage8u(sz)
arr = numpy.asarray(roi_im) # numpy view
arr[:, :] = 0
arr[YT, XT] = 50
return roi_im, result_validator_func
def test_fast_vs_slow(self):
h, w = 3, 5
shape = h, w # 3 rows, 5 cols
results = []
for func in [slow.do_bg_maint, ria.do_bg_maint]:
running_mean_im = 4 * numpy.ones(shape, dtype=numpy.float32)
hw_roi_frame = 5 * numpy.ones(shape, dtype=numpy.uint8)
max_frame_size = FastImage.Size(w, h)
ALPHA = 0.25
running_mean8u_im = numpy.empty(shape, dtype=numpy.uint8)
fastframef32_tmp = numpy.empty(shape, dtype=numpy.float32)
running_sumsqf = 16 * numpy.ones(shape, dtype=numpy.float32)
mean2 = numpy.empty(shape, dtype=numpy.float32)
std2 = numpy.empty(shape, dtype=numpy.float32)
running_stdframe = numpy.empty(shape, dtype=numpy.float32)
n_sigma = 2.0
compareframe8u = numpy.empty(shape, dtype=numpy.uint8)
bright_non_gaussian_cutoff = 255
noisy_pixels_mask = numpy.ones(shape, dtype=numpy.uint8)
bright_non_gaussian_replacement = 5
bench = 0
func(FastImage.asfastimage(running_mean_im),
FastImage.asfastimage(hw_roi_frame),
max_frame_size,
ALPHA,
FastImage.asfastimage(running_mean8u_im),
FastImage.asfastimage(fastframef32_tmp),
FastImage.asfastimage(running_sumsqf),
FastImage.asfastimage(mean2),
FastImage.asfastimage(std2),
FastImage.asfastimage(running_stdframe),
n_sigma,
FastImage.asfastimage(compareframe8u),
bright_non_gaussian_cutoff,
FastImage.asfastimage(noisy_pixels_mask),
bright_non_gaussian_replacement,
bench=bench)
results_order = (
'running_mean8u_im',
'fastframef32_tmp',
'running_sumsqf',
'mean2',
'std2',
'running_stdframe',
)
this_results = [locals()[name] for name in results_order]
results.append(this_results)
for i, (slow_result_arr, fast_result_arr) in enumerate(zip(*results)):
name = results_order[i]
if 0:
print name
print slow_result_arr
print fast_result_arr
print
assert slow_result_arr.shape == fast_result_arr.shape
assert numpy.allclose(slow_result_arr, fast_result_arr)
