I was trying to be careful to not state that laptop batteries what Tropos was doing with dual-band meshing was unique. It seems like a perfectly obvious and reasonable extension of meshing that's almost identical to how multi-homing works in a fixed wired network with many routes, some slower or more congested than others at any laptop battery given time. Rather, Tropos's commercial rivals are not pursuing that path yet, having bought into directional backhaul.

Racially intolerant white canadian cops and security and their henchmen claim to be despots; following parasitically in the footsteps of their american counterparts, and wilfully engage in their racial profiling of non-whites, in racial harassment of non-whites, and in racially dehumanizing attempts to racially harass non-whites through intimidating physically, mentally, and spiritually; portraying their racial hatred of non-whites through causing wilfull and dehumanizing disturbance to non-whites through using illegal wall-see-through technologies and audio-bugs on non-whites' homes; through listening and watching through the walls of non-whites' rented and owned homes, and through their internet and private telephones. The perpetuators of these evil deeds do this from their cars using illegal equipment slyly given to them by the unworthy cops, and then
accelerating their cars loudly and intimidatingly near non-whites' homes and driving intimidatingly in presence of non-whites on streets, making threatening u-turns, driving intimidatingly right up and over sidewalks when a non-white is on the sidewalk, and throwing their ugly bullying weight around, in their shameless acts of cowardice. It is all done slyly, supposedly smartly, however, they cannot fool all the people all the time. The cops also participate themselves to wail their sirens abusively everytime non-whites move and talk inside their rented and owned homes in daily routine living, in addition to having their henchmen, and often, using their non-white gutless henchmen in cars, transport, shopping centers, neighborhoods, etc, to commit these ugly harassing racially profiling
deeds at all times day and night. Using non-whites to engage in racial harassment of other non-whites is an obnoxiously evil sinister humanely disgraceful intelligent move of the whites well-known for their ugly divide and rule tactics through their non-white henchmen.
laptop battery
laptop batteries

somebody help me~!
Could you explain gaussian-weighted window?
How can we get image gradient and orientation?

Replying to Tingfan:

It's an interesting comment about human vision, but I'm not entirely sure it's correct. It's tough to be sure from your comment exactly how strong a claim you're making. But there certainly are plenty of cases where order does matter for word recognition - shape is one cue, but there are others, some of which do rely on the 'relative location of local features'.

The invariance discussed in the paper refers to complex cells in V1 (approximately gabor filters - far simpler features than characters). Invariances at higher levels is a VERY active and controversial research area, and we know very little about the neural mechanisms involved.

If I'm misinterpreting what you said, I'm sorry.

This paper deals with monochromatic images. In the conclusion section, they mention adding illumination invariance color descriptors. Has this been done yet, and did it help?

SIFT seems to be very dependent on the setting of parameters, from the amount of divisions in each octave to the ratio between principal curvatures and amount of bins and grid size of the descriptor histograms. How sensitive is performance to all these parameter choices; and how much tinkering is needed for each new problem application?

"We assume that the original image has a blur of at least \sigma = 0.5 (the minimum needed to prevent significant aliasing)"

Why is it so?

About ``local image descriptor'', the author used a gradient orientation histogram over a small region(4x4) rather than taking gradients for each individual pixel. This idea is inspired by the neuron structure for visual cortex.

I'm quite agree with this. We human beings do not distinguish the "relative location of features in a local area". For example: if 'algorithm' is typed as 'alogirhtm', most people can still recognize it. Because we memory the word by it's shape instead of the order of character. That's might be the reason why "CAPTCHA" works well to distinguish human from machines.

In figure 1, it shows each octave of scale space repeatedly convolved with the Gaussian. Do you know about how many times these convolutions need to be repeated, and/or how do you know when to stop the iterative process? Thank-you.

In the "Accurate keypoint localization" section it discusses two thresholds in relation to the location of the extremum x_hat, 0.5 and 0.3. If x_hat > 0.5 in any dimension then the extremum lies at a different sample point, and if |D(x_hat)| < 0.3, the extrema were discarded. Is there some reason for choosing these threshold values (obviously they worked) and are they to be held constant or are they to be tuned for each application?

The tutorial on local invariant features mentions how performance for each detector depends on the scene type. How are the types of scenes defined, and would it be possible for you to go into which descriptors are best for what type of scene?

From my observations, the SIFT code on Lowe's web site is an executable and all of the parameters are set. Can you go over some of the important parameters in the SIFT algorithm and how they affect detection? For instance, can you change the degree of rotation invariance?

The paper briefly mentions the approach of Schmid and Mohr that uses features that are inherently orientation invariant. It seems that in some cases where the dominant orientation of a patch is not well-defined, it would be more robust and probably faster to use such a representation. I would be interested to see a system that could choose which type of feature (orientation invariant, or sensitive with an implied transformation) to use depending on the appearance of the local neighborhood.

Tom, do you know anything about Schmid and Moore's approach and if it is competitive with SIFT?

What is an "octave of scale space" exactly?