Detail Man
Forum Pro
- Messages
- 17,490
- Solutions
- 11
- Reaction score
- 2,745
Can you provide any concrete bases for that thinking ? The power spectral densities displayed above significantly increase below around 5 Hz. Thus, we might expect that Exposure Times shorter than around 200 mSec might result in lower magnitudes of positional variations. However, such an assumption itself does not speak to the nature of the hand-tremor motions themselves.Thanks for the reference. I think that as the exposure gets shorter, the amount of randomness in the image caused by camera motion decreases.However, it seems that the results of a study where subjects attempted to hold an imaging device as steadily as they could for 2 Seconds time reveals what is clearly stated to be "mostly random":
Source: "Hand Motion and Image Stabilization in Hand-held Devices", Etay Mar Or, Dmitry Pundik.
From the text of "Hand Motion and Image Stabilization in Hand-held Devices", Etay Mar Or, Dmitry Pundik, IEEE Transactions on Consumer Electronics, Vol. 53, No. 4, November 2007 (where "Figure 2" below refers to the pair of time-domain plots appearing at the top of the graphic, and "Figure 3" below refers to the pair of time-domain plots appearing at the bottom of the graphic):
The results are summarized in Figs. 2 and 3. In Fig. 2, we present the mean standard deviation of the motion path as a function of time. The mean is over the 85 motions. The standard deviation is a good estimate for motion blur in most cases. Fitting the results to a power law:
σ[handmotion] (t) = A * t^(a)
yields for x-axis a=0.75 and for y-axis a=0.69. This result is close, yet higher, than that of a random walk. This is expected as the motions are not fully random, and contain some inertial part.
Do you have evidence supporting a hypothesis of "inertial non-randomness" (or something like that) ?
Quoting from the text cited above:
We first noticed that during one second of motion, the amount of rotation was negligible. The maximal rotation during a 1/96 Second period is 0.12 Degrees, for a full 1 Second it is 0.88 Degrees. In the rest of the analysis we ignore the rotation, as its effect is considerably smaller than that of the translation.
.
It could be - but with seemingly significant reductions in accuracy as camera-subject distance increases - as is indicated by the text quoted below. Rotational motion manifests on a level independent of camera-to-subject distance, however.
Source: "Image Stabilization Technology Overview", David Sachs, Steven Nasiri, Daniel Goehl
.
Angular to linear displacement calculations can be performed on 1-dimensional (angular) movements - but that is a mathematical fact which exists indepedent from the evident loss of accuracy involved.
.
I assume that you mean this. item (have yet to see one positive assessment by Adobe PS users).
My experiences attempting such surely feel "chaotic" ! I know nothing about "chaos theories" - but my thinking is that the propogation of such breezes may well be effectively random in nature - though (in a linear model) "filtered" by "system-responses" of relevant nearby physical objects).
.
It seems that one might reconsider assuming camera/subject-motion as being non-random in nature.
DM
Last edited:
