I have no experience with images. I have to detect simple object in static image. For example I have image like:
I want to detect edges and remove background. Just to compare them.
Something like this.
Do u have any solutions of this problem? Images have often white backgrounds.
I've just thought about detect edges, and take everything what they contains.
To segment out the shoe-
Anadptive Threshold to remove the smooth changing background.
Sobelx, which removes the apparent background line, which i assume is
common for images of this setup.
dilate, closing operation to separate out the shoe.
Find contours, bounding box etc as per your choice.
Do an additional threshold if you want to remove the shadow at the bottom.
Basically I'm reading vehicle license plates using tessract OCR, however despite being able to emphasise text easily enough via changing contrast, reducing noise an so on, some 'parts' of the vehicle remain on the image which does cause the OCR to throw bad results.
For example take:
I can change this easily enough, such as:
I'm looking to eliminate the edges off each plate, here's another example:
I could remove the edges using pixel manipulation algorithm, however I don't feel it's the right method, and would cause quite a lot of problems.
I've been using the following application to test various methods such as morphology and eliminating the unwanted data, so far I haven't been successful.
http://www.codeproject.com/KB/GDI-plus/Image_Processing_Lab.aspx
However someone with knowledge of this could use the application on the article above to achieve want I'm trying, so feel free to give it a try.
Thanks
Please try to use stroke width transformation concept.
This concept use to segment text from natural images.....
I already did such an algorithm. I just can say that it works great. The secret is, that you need to know that the light is coming just from one side perhaps. You cannot set the image to "black/white" just by using ONE threshold.
Detect the average luminance of parts of the image and use this luminance calculation to set the threshold for each region.
For example, if the left top is lighter, you need a lower threshold to make these parts not to bright. And if the bottom right has low light, you need to set the threshold higher to receive all existing light information.
Then, you need just to drive into the image from each side by using the method:
IsPixelAboveThreshold ?
If it is below, you are on the border, if it is above, you can say you are on the middle of the image with more brightness.
Regards
I have a windows forms application where I would like to find the handle of a DirectX application, find a DirectX button in that window, and Click it.
Is this possible, and if so, where is the best place to start?
Thanks!
This is really not possible.
Unlike Windows Forms, a "DirectX Button" has no distinct windows handle. You don't actually "click" a button in a DirectX window, but rather send a mouse click to the entire DirectX window, and allow the application to do its own hit testing to "click" the "button" (which is just a graphic).
The closest you could do would be to send a mouse click to a location within the window - however, finding the location of a "button" could be very difficult.
Actually, it isn't too difficult if that is the case. A DX object can receive an click if you can isolate the object with the mouse through a reversal system, depending if it is 3D-based or 2D-based. 2D is using masking effects, and 3D adds on to translate to and from 2D/3D based space on certain properties utilized in vector calculus.
Pretty much from logical space, in 3D, a slice of the 3D object is mapped to 2D space, which is then mapped to camera space, which is then rendered. Crude, there is much more to it then that, but I'm keeping it as simple as it can be. Now the task is to find the 2D volume/per-pixel location, translate a copy of the 2D render of a 3D object on an image that will be/is already a DeviceContext for the monitor to show () to a map, and unionize a clickable object (which will run the intended code, just as a button should) to the 2D map object copy.
In other words, find where in the screen the object is rendered and generate a clickable mask over the rendered portion of the screen. Whether it is a bitmap mask, or a polygon mask, or a cube mask, or any other shape you can mathematically create; so long as the 2D clickable object actually interfaces with the mouse position and state in an constant-update system.
For example, say that application is full screen, that the resolution is 800 x 600 (keeping it simple, it could be anything), and that the object you wish to be able to interact, that is rendered in a context that doesn't permit it of it's own accord (*1), and that the rendered object shows a rectangle. Lets say that is the point of it, and the system you use for the translation outputs a Rectangle (C#.NET code, yea, but it's just an example, there are better floating around). So we can say that the mask rectangle coordinates are identical to to the rendered object coordinates.
So let's say: "Rectangle mask = new Rectangle(100, 100, 200, 100); // X starting point, Y starting point, Length, Width." Pretty self-explanatory. Add a per-frame cursor position compared with mask positions, will indicate if the cursor is above the mask, and with extra coding, can indicated if the cursor clicked (or other events) while being within the mask's range.
So let's say the button opens the main menu. And an event happens, particularly a "MouseClick" event, while the cursor hotspot is at Point(125, 200). Which of course was in the range of the mask. In this instance, through all that trivial code, opens the main menu. It is easy to add new clickable objects within this system if it is done right.
Yes, other, easier ways exist. But from them all, this is one of the more dynamic and flexible systems. Uses range from the cash you get as loot in a game that you have to click to pick up, to the UI, to the menu's, to anything clickable. The range isn't in just games, either. Custom UI designers who like to reinvent the wheel can and have taken advantage of similar systems, such as myself, like to define every aspect from the microcode to the macrocode. Precisely, to define the macrocode as it is needed, as it is expected and without unnecessary bloat through microcode management. Not saying I like microcode, as it can be way too bloody finicky, and often times, is. Worth it for the few hundred FPS you get from streamlining it right (the first time), though... >:-D
*1 = (predictable in 3D Rendering models, from the point of view Rendering is not Modifying (which seems to be including object selection (through the mouse/cursor position and all that), which is essentially what your looking for) and as such not included (directly). Through translation it becomes possible, with the finest precision if done right. Further to that, DirectX, OpenGL, and others like, are defaultly Renderers.)
How is it possible to dynamically fill a container? Let's say to fill a big circle with small circles, recursively. Just fill the space fine.
I would like to use it for data hierarchy display.
To make it clear:
If you want something off the shelf, have a look at Graph#
http://graphsharp.codeplex.com/
videos here:
Simple usage scenario
http://www.youtube.com/watch?v=VTbuvkaPGxE
Data Visualization with Graph#
http://www.youtube.com/watch?v=agDPDzqB4o0&feature=related
It does dynamic graph layout and is fairly easy to use. There are a choice of layout algorithms [see sample app] each of which are configurable.
however filling until an area is "full" is not something that'll work out of the box. Although you could for example create a graph, lay it out & then measure the ratio between the size of a vertex and the size of the whole graph, then add or remove vertexes until you hit upon your desired density. I would hazard you could pretty quickly by trial and error come up with a quick and dirty forumula between the size of canvas to fill & the number of vertexes you should add.
Note that you can customise the vertex templates pretty easily to be any kind of data [this is standard wpf but Graph# specific examples can be found on http://graphsharp.codeplex.com/discussions ]
if you wanted to code your own layout you might like to have a look at some of the techniques Graph# use... for example a dynamic zoom component the source for which is available here: http://wpfextensions.codeplex.com/
hope that helps a little
I would like to develop a simple point&click game in C# with the default drawing libraries...no openGL/SDL/Tao for this project.
Suffice to say, I am curious as to the best ways to draw clickable images in layers on a form.
Ideally, I would have
1) Environment layer (pathways, doors, etc)
2) Object layer (items)
3) Character layer (enemies)
Ideally, the layers beneath other layers would still be viewable, so I could still see the environment underneath an object (so whatever component I use to draw the object to the form needs to be transparent).
This game is going to be tile based...so I will be generating a 2D array of some sort of component and putting those onto the form. The question is, what component should I use? A friend has recommended to generate panels and drop those onto the form and use the background image property, but is there a better way?
I know that this is not the ideal way to develop...this is more of a prototype for myself. Later on I will probably move it to Tao if I get anywhere, but for now (ie, the next year or so), I would like to keep it extremely simple.
Tiles are simply logical chunks with which to organize your UI. You do not need to necessarily have tiles in the form of Panels to use the "tile" idea. Switching tiles can simply mean that your character has reached the right edge of tile A and you will therefore draw tile B and place the character on the left edge of that tile. So your memory structure should be a 2D array of logical tile objects that contain information about what to draw but not necessarily Panels, though that is an option, i don't recommend it.
Layers: You don't really need to worry about layers. All you have to do is draw in the correct sequence: Environment first, followed by objects and then characters. .Net and probably all frameworks paint back to front, which means that things drawn first may be entirely or partially obscured by objects drawn later. The "layering" happens automatically. If you just draw your objects within their own bounds then you do not need to worry about transparency. E.g. a 20x20 pixel character is drawn within a 20x20 pixel bound.
This is all assuming raw e.Graphics.Draw* and e.Graphics.Fill* calls in the Paint event handler of course. As with anything in software, there are 101 ways...