Imagine having an array of images like these
Background is always white(even in the 3rd pic, main object there is that big brown rectangle with shapes inside)
No matter of given type of the image you would need to:
1) find main object boundary rectangle
2) crop it out like this
3) and place it in the center of a blank square image.
How would you achieve this? I already know how to crop out anything knowing rectangle and place it anywhere but I just need to know which way would be the best to make the 1st step.
Vision API can detect rectangles, faces and barcodes, but it seems what I need is even more simple.
I just need to find leftest, rightest, top and bottom non-white pixels and it will be my bounds.
Is there any way except iterating pixelBuffer for each pixel?
What is the type of these images? UIImage? CAShapeLayer? In most cases, you should be able to get the .frame from each image in the array, which will give you a CGRect the X and Y origin coordinates, as well as height and width dimensions. You should also have access to .midX and .midY coordinates, or .center.x and .center.y to find the midpoint you're looking for. Unless what you're talking about is taking in a flattened bitmap like a .jpg or .png and running some shape detection on the contents, in which case you would need something like Vision to accomplish what you're trying to do.
Related
i have the requirement to perform a redaction in itext7. We have several rectangles which have been selected by the user. Some of these have been rotated. I have not found the ability to rotate rectangles in itext7. Usually, how we draw "rotated" rectangles is to perform some mathematical operations on a "fake" rectangle we draw in the code, and then draw them either using a series of lines like so:
if (rect.mRotation > 0)
{
r.Rotate(DegreeToRadian(rect.mRotation));
}
c.MoveTo(r.TopLeft.X, r.TopLeft.Y);
c.LineTo(r.TopRight.X, r.TopRight.Y);
c.LineTo(r.BottomRight.X, r.BottomRight.Y);
c.LineTo(r.BottomLeft.X, r.BottomLeft.Y);
c.LineTo(r.TopLeft.X, r.TopLeft.Y);
c.Stroke();
In the case of images, or something similar, we are unable to do the above. In this case we use an affinetransform to simulate the movement, which is applied to the image before it is added to the document. Both of the previous methods work perfectly.
Unfortunately for us, the pdfSweep tool only accepts (iText.Kernel.Geom) rectangles. We are looking for a way to be able to still pass an iText.Kernel.Geom.Rectangle which has had transforms applied (ie. a rectangle which has been rotated). We have tried setting the llx/urx values manually using the setBBox method, but this wont affect the rotation.
Does anyone know how we can go about redacting items over a given rectangular area which has been rotated?
Thanks
One TV screen recognition project, i need to clip the TV Screen from one image.
The TV screen actually is rectangle. But It's obvious that the TV screen is out of shape in the image from phone camera. My question are:
How to detect the any 4 sides polygen(it's not rectangle) in the image.
After i know the polygen area on the image ,how to retrieve the area to Mat.
After solve quest2, How to convert the Mat of 4 sides polygen to rectangle Mat which is fixed W/H radio.
It's very helpful that give some code sample to reference.
Thanks your answers!
if you want to detect the edges of your TV screen you can use some border
detection (like Canny) and then use Hough transform to obtained the lines.
If you then extract the points corresponding to the intersection of the lines
you can create an homography matrix H (3x3). Finally, using this homgraphy you can
"deform" your original image to a reference frame (in our case the rectangle
with a given aspect ratio). The homography is a transformation from plane
to plane, so it's exactly what you will need here.
If your going to use OpenCV (which is always a good choice!),
here are the functions that you could use:
Canny() - find edges in the image
HoughLines() - detect lines
findHomography() - this function finds from a set of correspondances,
the homography matrix. In your case, you will need to pass the method
as 0.
warpPerspective() - the function that your going to use to "deform"
the image to a reference frame.
Obviously, you can find similar functions for MATLAB and others...
I hope this helps you.
I would like to extract the white areas/bright areas of an image and place a custom objects in those areas. I need to know which framework to work with. if anyone has done something similar, I would appreciate an answer. I know how to get Pixel values, however the hard part is creating a Bloom/star effect in those highlighted areas.
You could make a mask where the luminance value was above a threshold, then blur or whatever the mask and composite above the image.
I want to use bitmap images as a "map" for levels in iphone game. Basicly it's all about the location of obstacles in the rectangular world. The obstacles would be color-coded -- where the white pixel is, there's no obstacle. Black means there is one at this point.
Now I need to use this data to do 2 things: (a) display the level map, (b) for in-game calculations. So, in general, I need a way to read the data from the bitmap and create some data structure (matrix-like) with those information - to both overlay the bitmap onto the level graphics as well as to calculate collisions and such.
How should I do it? Is there any easy way to read the data from image? And what's the best format to keep the images for this?
Have you looked at how Texture2D translates an image file to an OpenGL Texture ?
Tip: take a look at this Method in Texture2D.m:
- (id) initWithCGImage:(CGImageRef)image orientation:(UIImageOrientation)orientation sizeToFit:(BOOL)sizeToFit pixelFormat:(Texture2DPixelFormat)pixelFormat filter:(GLenum) filter
In 3D apps, it's quite common to use this kind of representation for height maps, in a height map, you use a Texture with colors that range from black to white ( white represents the maximum altitude )
For example, from this:
To this:
That was just to tell you that your representation is not that crazy :).
About reading the bitmap, I would also recommend you to read this (just in case you want to go deeper)
Hope I helped a bit!
I'm having an issue with drawing to areas outside of the MKMapRect passed to drawMapRect:mapRect:zoomScale:inContext in my MKOverlayView derived class. I'm trying to draw a triangle for each coordinate in a collection and the problem occurs when the coordinate is near the edge of the MKMapRect. See the below image for an example of the problem.
In the image, the light red boxes indicate the MKMapRect being rendered in each call to drawMapRect. The problem is illustrated in the red circle where, as you can see, only part of the triangle is being rendered. I'm assuming that its being clipped to the MKMapRect, though the documentation for MKOverlayView:drawMapRect makes me think this shouldn't be happening.
From the documentation:
You should also not make assumptions that the view’s frame matches the bounding rectangle of the overlay. The view’s frame is actually bigger than the bounding rectangle to allow you to draw lines for things like roads that might be located directly on the border of that rectangle.
My current solution is to draw objects more than once if they are in a maprect that is slightly larger than then maprect given to drawMapRect but this causes me to draw some things more than needed.
Does anyone know of a way to increase the size of the clipping area in drawMapRect so this isn't an issue? Any other suggestions are also welcome.
I ended up adding a buffer to the rect passed in to drawMapRect:mapRect:zoomScale:inContext and using that to determine which objects to draw. This results in more objects being drawn than needed, but not by much.