|
Hello, I have considered doing following project:
Imagine that you have simple graphics (like some schematics, logo, etc). I'd like to investigate some other approaches to saving the image like traditional bitmap/vector files.
It would use some kind of advanced algorithms to generate set of graphical operations (draw line/shape/curve, fill shape, draw text somewhere) that would be drawn to some bitmap. The program would then evaluate how much the generated image look like the original (like fitness function in GA), I mean, evaluate many solutions, find the best ones, combine/mutate them in some way to produce better results in the next generation.
It would not be saving the image in the true sense of word, it would be more like finding algorithm on how to recreate it with the best precision using elementary graphical operations. Maybe it would be useless, maybe not.
My question is, *how* should I start? I read something on the topic of genetic algorithms, but most of the articles dealt with using binary string (genes) for representing the operations (but wouldn't it be too long to encode many (tens, hundreds) graphical operations with parameters?) I mean, if we had about hundred operations (8-bit identifier) with average of four 16-bit parameters, it would be about 100*8+100*4*16=7200 bits of information to combine, mutate and eventually, evaluate.
I know it will be painfully slow, to have some kind of fitness evaluation function that would need to draw image from the "genes", but this would be a research project. I am just wondering if something like this is possible (I think it could be), and if yes, someone could please point me in the right direction to start.
thanks,
Juraj
|
|
|
|
|
G'day Juraj,
Juraj Borza wrote: The program would then evaluate how much the generated image look like the original (like fitness function in GA), I mean, evaluate many solutions, find the best ones, combine/mutate them in some way to produce better results in the next generation.
You might take a look at Bezier curves and spline functions. In the early nineties i saw a demo of a graph of a teapot on a Unix workstation. The image was entirely generated by means of splines. The light reflection is very realistic, since the first order derivatives of the splines are continuous, at the edges as well.
Regards,
Stan
|
|
|
|
|
It sounds to me like you are conceptually describing a compression algorithm. But I think you are really looking for a different result. Can you clarify your goal for this algorithm?
Otherwise, I suggest you look at using wavelets to represent discrete sections of the image. The wavelets are easily reconstructed to a reasonable representation of the image. Depending on how you structure them, they can represent it perfectly, or in as much abstraction as you desire.
If you continue to do the same things you always did,
don't be surprised if you get the same results you always got.
|
|
|
|
|
Your instincts are correct: This is feasible and isn't that hard. You need to define your "genes" and mutation operators. Genes will be more than single bits, and you may want to restrict recombination to gene boundaries.
Traditional genetic algorithms use a fixed number of genes, but you will want a variable number, to represent different approachs to drawing the image. Google "genetic programming" to see genetic-algorithm examples with variable-sized genotypes.
There's already a similar method for compressing images called "Fractal Compression", in which fractals are combined to mimic the image being compressed. It's more systematic than a genetic algorithm, and compression ratios on the order of 10,000 to 1 have been reported.
|
|
|
|
|
Suppose I have an image of a region of earth of arbitrary size and I have Lat/Long from two points on the map is there a simple formula for converting a given pixel to a Lat/Long that takes into account the curvature of Earth? Assume perfect lens and that the map is of such a scale and size that a given pixel distance will be the same across the image.
And yes, I know, if I could afford ARC/GIS it will do it for me.
Need a C# Consultant? I'm available.
Happiness in intelligent people is the rarest thing I know. -- Ernest Hemingway
|
|
|
|
|
Ennis Ray Lynch, Jr. wrote: Suppose I have an image of a region of earth of arbitrary size and I have Lat/Long from two points on the map is there a simple formula for converting a given pixel to a Lat/Long that takes into account the curvature of Earth? Assume perfect lens and that the map is of such a scale and size that a given pixel distance will be the same across the image.
Not positive from how the question is asked (by "image of a region" do you mean a rendering of the earth in perspective, ortho, a map projection?), but here is is the lowdown:
If you have a perspective image of the earth, find the local xyz coordinates, convert to ECEF, then to lat,long, alt.
If you have an ortho image, again depending on your projection, you get a local coordinate and convert through.
if you have a map projection, use the appropriate conversion to/from to get to ECEF and/or directly to lat/long.
http://www.gdal.org/[^]
http://www.maptools.org/[^]
http://www.bluemarblegeo.com/products/geocalc.php[^]
_________________________
Asu no koto o ieba, tenjo de nezumi ga warau.
Talk about things of tomorrow and the mice in the ceiling laugh. (Japanese Proverb)
|
|
|
|
|
I have a picture taken of the ground of which I know a few lat/long coords. I would like to move the mouse over the picture and have the mouse x,y translated into lat/long.
Need a C# Consultant? I'm available.
Happiness in intelligent people is the rarest thing I know. -- Ernest Hemingway
|
|
|
|
|
Ennis Ray Lynch, Jr. wrote: I have a picture taken of the ground of which I know a few lat/long coords.
if you don't know the location of the camera, or the projection of the image, you will have to do interpolation estimation based on the deltas between your lat/long points.
_________________________
Asu no koto o ieba, tenjo de nezumi ga warau.
Talk about things of tomorrow and the mice in the ceiling laugh. (Japanese Proverb)
|
|
|
|
|
However, interpolation between delta's itself will not work against longitude as the distance between two lines of longitude change based on the latitude. I could break out my trig but I just wanted to know if anyone had a good formula handy.
Need a C# Consultant? I'm available.
Happiness in intelligent people is the rarest thing I know. -- Ernest Hemingway
|
|
|
|
|
Ennis Ray Lynch, Jr. wrote: interpolation between delta's itself will not work against longitude as the distance between two lines of longitude change based on the latitude.
that is why they have map projections. GDAL does the Proj.4 conversions between formats. Generally you go through a cartesian system, like Earth Centered Earth Fixed (ECEF). http://en.wikipedia.org/wiki/Geodetic_system[^]
_________________________
Asu no koto o ieba, tenjo de nezumi ga warau.
Talk about things of tomorrow and the mice in the ceiling laugh. (Japanese Proverb)
|
|
|
|
|
The would require the image to be of a known projects. However, an image with lat/long points should be able to provide enough information to both project and translate itself. Like I said, I will just do the trig myself unless someone has an algorithm.
Need a C# Consultant? I'm available.
Happiness in intelligent people is the rarest thing I know. -- Ernest Hemingway
|
|
|
|
|
Ennis Ray Lynch, Jr. wrote: The would require the image to be of a known projects. However, an image with lat/long points should be able to provide enough information to both project and translate itself. Like I said, I will just do the trig myself unless someone has an algorithm.
Not necessarly... you simply take the existing projection with lat/long and come up wth the perspective information to co-reference the image points to lat/long, you can then use lat-long conversions through known projection types to come up with measurements and other projections.
This is how you do ortho-rectification of an image. Once you establish the co-referencing of lat/long for the image you stretch it to match a geodetically ortho-rectified image. The result is a bilinear interpolated re-projection of the orignal images, but in ortho projection. Most algorithms for orthorectification use about 6 points to get a good "feel" for the referenced information in the original. You can use the same information to provide any set of geodetic inforation since you have essentially created your own projection by definining the curves across the image in lat/long space.
_________________________
Asu no koto o ieba, tenjo de nezumi ga warau.
Talk about things of tomorrow and the mice in the ceiling laugh. (Japanese Proverb)
|
|
|
|
|
That, I think, may help me. Thanks.
Need a C# Consultant? I'm available.
Happiness in intelligent people is the rarest thing I know. -- Ernest Hemingway
|
|
|
|
|
Ennis Ray Lynch, Jr. wrote: That, I think, may help me. Thanks.
then pretty much any curve fit between lat/long to x/y will work. The more points you have to come up with the curve of image xy to lat/long the better. Then you can look up any point map its lat/long with the appropriate interpolated position and come up with its lat/long
_________________________
Asu no koto o ieba, tenjo de nezumi ga warau.
Talk about things of tomorrow and the mice in the ceiling laugh. (Japanese Proverb)
|
|
|
|
|
i have some clue, which i already implemented, iam not sure weather it works for u or not.
1.you have the graticules(longitude and latitudes) of two pixels right?
2.if so, you can caluculate two factors i.e. ppsX (pixels per second along X) and ppsY. i think simple mathematical formulas related to co ordinate geometry has to be implemented.
3.now, based on that, you can caluculate graticule of any pixel.
any more support required? you are welcome.
--------------------------------------------
Suggestion to the members:
Please prefix your main thread subject with [SOLVED] if it is solved.
thanks.
chandu.
|
|
|
|
|
that's only valid on scales small enough that the deviation between real life and a Mercator projection isn't significant.
--
If you view money as inherently evil, I view it as my duty to assist in making you more virtuous.
|
|
|
|
|
I want to figure out the formula in getting the following:
$2500 loan amount
239.9961% APR
12 month term
Bi-weekly payment (26 payment total)
What is the formula for each payment?
What is the formula for the Total Finance Charge?
How can I calculate Bi-weekly equal payments??
We know that each payment will be $251.96 (Equal Payments) But how have they come up with such formula?? And where?
HELP!?!?!?!?!
Victoria
|
|
|
|
|
will (p*t*r)/100 work for you?
or you may have to just fine tune this basic formula.
if you can elaborate your requirement more, then i can explain with better clarity.
|
|
|
|
|
The only formaula that seems to work is one over 364 days.. Here is it:
=PMT(APR/364 x Number of payments x 1 x Amount finaned)x -1 - APR
But shouldn't it be 365 days? 
I did this on excel and that's all I can come up with. Let me know what you think.
Victoria
|
|
|
|
|
Vtoria wrote: HELP!?!?!?!?!
NO KIDDING! Who in the world is charging 240 APR interest??? 
BTW, at 240 APR, $252 won't even make the interest payment, let alone pay it off in a year! 
David
|
|
|
|
|
Hi David,
This is for an installment loan that is VERY high risk. No credit check, or anything like that. But that isn't the issue at stake. The issue is how are they coming up with the formula for equal bi-weekly payments at that APR?. To answer your question, YES, it does pay it off:
monthly principle interest balance
251.96 70.64 181.32 2,429.36
251.96 27.72 224.24 2,401.64
251.96 30.27 221.69 2,371.37
251.96 33.07 218.89 2,338.30
251.96 36.12 215.84 2,302.18
251.96 39.45 212.51 2,262.72
251.96 43.10 208.86 2,219.63
251.96 47.07 204.89 2,172.55
251.96 51.42 200.54 2,121.13
251.96 56.17 195.79 2,064.97
251.96 61.35 190.61 2,003.61
251.96 67.01 184.95 1,936.60
251.96 73.20 178.76 1,863.40
251.96 79.96 172.00 1,783.44
251.96 87.34 164.62 1,696.11
251.96 95.40 156.56 1,600.71
251.96 104.20 147.76 1,496.50
251.96 113.82 138.14 1,382.68
251.96 124.33 127.63 1,258.35
251.96 135.81 116.15 1,122.54
251.96 148.34 103.62 974.20
251.96 162.04 89.92 812.17
251.96 176.99 74.97 635.17
251.96 193.33 58.63 441.84
251.96 211.18 40.78 230.67
251.96 230.67 21.29 0.00
Victoria
|
|
|
|
|
OK, I missed the point about 26 payments. I thought you meant 12 monthly payments.
|
|
|
|
|
Use the same formula I provided earlier, but substitute 2.4/26 (the biweekly rate) into the formula for 'r' and 26 for 'n'.
It gives a different answer than 251.96, but I see that your first principle amount is very high in comparison, indicating that your first payment is being made before the full 2-weeks have expired. This will change your payment amount.
David
ps. is it really legal to charge that much interest where you live? I've heard of such things, but I think it's a disgrace. These kinds of loans are specifically designed to rape the poor.
|
|
|
|
|
Ok.. I'll try it.
Yes, it is legal, in most states. I can see where some people whould call them loan sharks, but I also see their side to it. These are companies that do not do any type of credit check. All they require is a valid drivers license, current paystub, ative checking account and phone bill. Who knows what they have out there. If I was going to lend someone $1000 who had a track history of late payment, I would want $1500 back in 6 months. These types of loans are so risky! Especially because more than half have bad credit. Now, let't say someone needs a loan until their bonus checks comes in in 2 weeks, and they need to make rent, car payment and extra money for food. (We all know that a lot of people live check to check). They take a loan for $1,000; in 2 weeks they pay $1,090.00! So it cost them $90.00 to take out a 2 week loan! NOT BAD. And it kept them from being negaitive in their accounts or get charged NSF fees, etc.. I say short term, it's worrth it. Long term.. NOT AT ALL!
Victoria
|
|
|
|
|
There's been exactly *one* actual study on this sort of lending. With the cooperation of a South African bank they took a group of people who were just under the approval threshold for a loans with average values of ~1000, and 4 monthly payments of ~360 (don't recall if values were in US or local currencies) and granted approval to half of them anyway. They then did a 6mo followup on both groups and found the ones that had been granted the loan were on average better off financially than those that had been denied. I read about it in a recent issue of The Economist, but since I don't save back issues I can't tell you which one it was.
--
If you view money as inherently evil, I view it as my duty to assist in making you more virtuous.
|
|
|
|
|
General 
News 
Suggestion 
Question 
Bug 
Answer 
Joke 
Praise 
Rant 
Admin 
Submit an article or tip
