We have a system where people are being taken a face shot via a DSLR camera. We need the people's images with transparent background. What we're currently doing is taking the image and editing and cropping it in Photoshop, removing the background image with the Magic Eraser tool.
What I am looking for is a way to parse the image and automatically erase the semi-white background we have, along with the resizing and cropping. Is there some kind of library or code sample that does this without requiring manual intervention?
This is a real complex problem. Like the answer below suggested you'll need to do a fuzzy match on each pixel and set it to be transparent but you also need to detected other nearby pixels to make sure they are not close in color. A white tag on the shirt, white eyelids, hair, pale skin reflecting the flash. All are candidates to be removed by any greedy fuzzy logic.
Think about the Magic Wand tool in Photoshop. How good is it at detecting the edges of the person in the picture? Yeah, and that's the top standard of image editing software with thousands of engineering hours behind it.
This is not a feasible request for a Q&A format, and this is one of those things that humans just do better than machine. BUT, that doesn't mean it's not possible, and who knows, you might be the one to do it. Just don't do it in VB.NET please :)
Some pseudo-code to get an idea of what you need to do:
Bitmap faceShot = Bitmap.FromFile(filepath)
foreach pixel in faceShot
//the following line is where the magic happens, you can do any fuzzy match on the color that suits you
//figure out your color range and do a fuzzy match percentage wise
if (pixel between RGB(255,255,255) and RGB(250,235,215)) //white and antique white
pixel.setAlpha=0
endif
end foreach
You could start with this as a starting point for processing a single image,
http://www.java2s.com/Code/VB/2D/ProcessanImageinvertPixel.htm
Basically, if you have a constant background color (like the TV green-screen), it's just a matter of selecting pixels close to the color you are erasing and setting their Alpha level to 0 (transparent). Treating the RGB values like XYZ coordinates, you can do a 3d distance from your background color, and make everything within a certain threshold transparent.
As an improvement, you could also make everything within another threshold semi-transparent so the edges right around hair and stuff like that look softer and less harsh.
Alternatively, you could probably do the same exact thing with good results in Photoshop, as it should support batch processing.
Edit, thinking about it some more, you may want to use a green screen type background as well instead of an off-white one like you stated, as you may make people's eyes transparent. I would definitely try to batch it in Photoshop/Gimp/etc.
Related
I'm using the EMGU CV .NET library. I noticed that when I take pictures of anything with color, the colors usually get "washed out" if the background is dark(ish). General rule of thumb I've found is that, the darker the background is, the more washed out the colors are.
Here is how I'm retrieving the image from the camera with EMGU.
Dim imgFeed As Bitmap = mCamera.RetrieveBgrFrame.ToBitmap
In the images below (cropped out some of the background on both), the left image is on dry white cement and the right image is on wet white cement. You can see the "washed out" color especially on the first tag, which is bright orange duct tape.
Here is another image, taken on black pavement in the sun, which in reality is much darker than the white cement, but appears similar in color to the background in the wet cement image above.
Is there some sort of auto-balancing that's occurring in the EMGU library? If so, can I stop this from happening? I need to see the colors more clearly than the background. I've read about _EqualizeHist() and I implemented it, but that did not help me see the colors any more clearly; adding contrast to the image didn't really help because the colors were already close to white.
Update
After reading Spark's answer, I found the SetCaptureProperty() method. I see that you can disable the auto exposure property by setting the value to 0 as shown below.
mCamera.SetCaptureProperty(CvEnum.CAP_PROP.CV_CAP_PROP_AUTO_EXPOSURE, 0.0)
Sadly though, with the particular camera I'm using, it looks like the driver does not support changing this property.
This is nothing to do with the algorithm. It is the behavior of Auto Exposure (AEC) algorithm running inside camera chip. Try disabling auto exposure control of camera and reduce the manual exposure level.
Little theory: Most of the AEC algorithm works with full frame weighted method. So in the sample which you showed for white washed case, the black background occupies more portion of the image, which make AEC algorithm to assume that the image is too dark and hence increase exposure level internally.
I'm trying hard to nicely blur a red circle but everytime i get gradient levels of red and the image looks choppy.
Before:
http://i.imgur.com/6yzMhFI.png
After:
http://i.imgur.com/2dZl4ph.png
How i can acheive a smooth blur ?
If you are referring to the visible circles that separate the gradation levels, that is called banding Here are some ways to fix that:
Increase your document's bit level from 8-bit to 16-bit
This will increase the amount of colors your file can represent, creating more colors that can be used to represent the gradient, making it smoother in appearance.
In Photoshop navigate to Image>Mode>16-Bits/Channel
In GIMP 2.10 (or higher?), navigate to Image>Precision>16 bit..
Display or system settings might be unable to display enough colors
If changing the bit depth does not fix the issue then you might have a hardware or system settings issue.
If it's a hardware issue, your monitor might not have the capability to display enough colors to render the gradient smooth
If it's system settings you will need to go to your operating systems color depth setting, usually located under the system's display settings. It could say something like Millions of Colors, or True Color (32-bit).
The last thing related to settings is that you have a bad color profile set in your system or in your image editing software. It's beyond the scope of this answer. If you don't know how to color calibrate your monitor, then it most likely isn't this and you can skip this.
If you have to have 8-bits
If you absolutely have to keep your document in 8-bit color space then you will have to use dithering or add some noise to your image to confuse the viewers brain into seeing a smooth gradient.
Noise or dithering will confuse the viewers brain into seeing a smoother gradient by setting some focus on the imperfections of the noise/grain/dithering. This doesn't exactly answer your question, but it is about the only option you have if you keep your ultra smooth gradient in 8-bit mode.
Good Luck!
I think you are applying the Gussain-Blur to the entire image try to Select the red circle and apply the Gussain-Blur filter to it
Suppose there there is a scene as follows:
There is a scene with the same size as the frame of the device. The scene has a red ball, which is able to move throughout the 'world'. This world is defined by black and white areas, where the ball is ONLY able to move in the area that is white. Here is a picture to help explain:
Parts of the black area can be erased, as if the user is drawing with white color over the scene. This would mean that the area in which the ball can be moved is constantly changing. Now, how would one go about implementing a physicsBody for the an edge between the white and black areas?
I tried redefining the physicsBody every time it is changed, but once the shape becomes complex enough, this isn't a viable solution at all. I tried creating a two-dimensional array of 'boxes' that are invisible and specify whether most of the area within each box is white or black, and if the ball touched a box that was black, it would be pushed back. However, this required heavy rendering and iterating over the array too much. Since my original array contained boxes a little bigger than a pixel, I tried making these boxes bigger to smooth the motion a little, but this eventually caused part of the ball to be stopped by white areas and appear to be inside the black area. This was undesired, since the user could feel invisible barriers that they seemed to be hitting.
I tried searching for other methods to implement this 'destructible terrain' type scene, but the solutions that I found and tried were using other game engines. To further clarify, I am using Objective-C and Apple's SpriteKit framework; and I am not looking for a detailed class full of code, but rather some pseudo-code or implementation ideas that would lead me to a solution.
Thank you.
If your deployment target is iOS 8, this may be what you're looking for...
+ bodyWithTexture:alphaThreshold:size:
Here's a description from Apple's documentation
Creates a physics body from the contents of a texture. Only texels
that exceed a certain transparency value are included in the physics
body.
where a texel is a texture element. You will need to convert an image to the texture before creating the SKPhysicsBody.
I'm not sure if it will allow for a hole in the middle like your drawing. If not, I suspect you can connect two physics bodies, a left half and a right half, to form the hole.
I have a layer with a sprite of a simple black donut. I want the user to be able to draw on the sprite in a different color (which I've managed to do without any problem using CCRenderTexture).
My question is how I can calculate whether the image has been traced at least 95% (meaning, find out when 95% of the black pixels are now the new color). I've tried methods like taking a screenshot of the layer and counting the number of black pixels, but it hasn't worked that well (using this solution: https://stackoverflow.com/a/1262893/1577738).
It would be even better if I could just change the color of each pixel as it's touched (to avoid issues with coloring out of the lines). I could theoretically just split the donut into like 10 sprites and change that section's color if the user touches it, but that seems ridiculous if I give the user options to use a bunch of different colors.
Am I going about this the wrong way? Your suggestions are much appreciated!
Reading pixel colors will be rather inaccurate and slow. I suggest dividing the area into smaller rectangles (ie 8x8 or 4x4) and then flag each as "visited" when the user draws on it. If most rectangle areas are flagged, the user has drawn on most parts of the texture.
i would like to program a little app that will change the colors of the screen. im not talking about the darkness. i want it to mimic what it would look like if for example you put on blue lenses or red lenses. so i would like to input the color and i want the screen to look as though i put on lenses of that particular color. well i actually need the program to semi-permanently change the users experience on the computer. i need the computer for the entire session that it is turned on to be changed this color
Transparent, Click Through forms might help you out. It makes a nice see through form that lets mouse clicks pass through it. The solution is in VS2003 format, but it upsizes to 2008 nicely. You could take that sample, rip the sliders off, get rid of the borders and make it fullscreen + topmost. I don't know if it'll accurately simulate a lens though, someone more into optics can tell me if I'm wrong :-)
If the lenses you are trying to simulate are red, green or blue, simply zeroing the other two colour components of each pixel should work. A coloured filter lens works by passing only a certain wavelength of light, and absorbing the others. Zeroing the non-desired components of the colour should simulate this accurately, I believe.
To simulate cyan, magenta, or yellow lenses, zeroing the one other colour component (e.g. the red component in the case of cyan tinted glasses) should work.
I'm not sure how to generalise beyond these simple cases. I suspect converting to say HSV and filtering based on the hue might work.
To change this for the entire system and use it in interactions with ordinary programs, you could change the colour profile for the display. For paletted/indexed-colour displays, this could be done by changing the colour look-up table (CLUT) for the display adapter. PowerStrip is a handy utility with versatile colour controls that should be able to achieve this quickly and easily on modern display adapters (e.g. by adjusting the red, green and blue response curves independently).
I came across Color Oracle and thought it might help. Here is the short description:
Color Oracle is a colorblindness simulator for Windows, Mac and Linux. It takes the guesswork out of designing for color blindness by showing you in real time what people with common color vision impairments will see.
Take a snapshot of the screen, convert each pixel into its grayscale value, then change the pixel value to a percentage of red. This will preserve the contrast throughout the image while also presenting a red tone.
To convert to grayscale in C#:
https://web.archive.org/web/20141230145627/http://bobpowell.net/grayscale.aspx
Then, to convert to a shade of red, zero out the values in the green and blue for each pixel.
(You can probably do the above in one shot, but this should get you started.)