I am collecting frames from Camera and trying to do some image processing and OCR on them. When the camera is perpendicular to the target, i have no problem with cropping and image processing. But when i try to capture the frame with angle, the target expectedly turns into a Trapezoid rather than rectangle.
I am using OpenCV for all the image processing stuff. I was wondering , what can i do to get the area that i want when its trapezoid?
For example Can i somehow transform that trapezoid image to rectangle with OpenCV?
Thanks
Your problem is called keystone correction:
http://en.wikipedia.org/wiki/Keystone_effect
Look at the similar question:
Executing cv::warpPerspective for a fake deskewing on a set of cv::Point
Related
I have a fragment in which I'm using a TextureView, and I'm using the following link as a reference:
https://github.com/googlesamples/android-Camera2Basic
Is there a way to modify the scope of the camera to take a "landscape" (it's not really a landscape, just a different scope) picture, even though I'm on portrait mode?
I'm attaching a photo of what I'm trying to achieve. I have a round white frame, I want to the scope of the camera/TextureView to be in that frame, and I want to add a button that takes exactly what's in that frame (with rounded corners). Is that possible?
It is feasible.
You can use opengl to achieve it.
First, draw camera frame to a external gl texture, then draw the frame in the gl texture to the screen.
Crop the frame during the process of drawing to screen with modifying the texture coordinate and you will get your effect.
the way is here:
Crop video before encoding with MediaCodec for Grafika's "Continuous Capture" Activity
And the round corner effect can be also implement through modifying the texture coordinate. the way is here:
How to make TextureView play video with round corners and bubble effect
I want to display a camera preview in a circular shape using the camera2 api. I want to display the preview in a circular shape, but I dont't want the image to be captured in a circular shape.
The captured image would be a face( later want to implement face detection and auto capture). I did have a look at few questions already asked, but none of them are with the new camera2 api's and most of them talk about having an overlay image cropped with a transparent circle. But this will not work in a case where I need to auto detect a face(as the face may appear out side the cropped circular image).
Is there any way I can implement this ? I did try an example with TextureView and set it to a LinearLayout with fixed width and height, but the preview appeared a bit squeezed and in a square shape.
I don't see why face detection matters here - if you enable the camera API's face detector, it'll run on the full image no matter what you do in drawing it inside a circle.
You can either use a circle overlay on top of a correctly-shaped TextureView or SurfaceView, or do your own OpenGL rendering of a circle with the camera preview as a EGL texture.
The latter you'll probably want a GLSurfaceView for the OpenGL drawing context, and a SurfaceTexture to send camera data to and expose it as a EGL texture.
JPEGs captured will still be full-FOV, and the camera API will know nothing about your circular preview drawing, so face detection and everything else will work on the full field of view.
In my android project I want to add an overlay (which is simple image with text on it) on top of the camera view.
The overlay is not simple rectangle but it is transformed at each camera frame.
What I'm looking for is kind of like image below:
I'm getting camera frame with OpenCV library (CvCameraViewListener and CameraBridgeViewBase).
My question is what is the best and fastest way to do this?
How can I transformed the overlay at each frame in the way as figure above shows.
Any help and suggestion is appreciated.
I have user tesseract ocr for my android project to recognize text from an image taken from the camera. But the results are not accurate. I want to optimize the image using opencv. I want to achieve the following for the captured image which is decoded in Bitmap.Config.ARGB_8888 format:
Detect the objects in the resized image.
Once the object is identified, compute its border w.r.t original image. (This is for removing the camera angle effect)
Extract the object from original image, by applying perspective transform.
Apply white balance to remove lightening effects.
In the example provided by with the tess_two api, they are using Leptonica for the image manipulations like drawing the bounding boxes around the words..But in my case I want to use OpenCV...Your guidance will be highly appreciated...
That's a lot you are asking for, and depending on the object may be impossible. You should check out the tutorials on 2D feature detection and object detection (http://docs.opencv.org/doc/tutorials/features2d/table_of_content_features2d/table_of_content_features2d.html and http://docs.opencv.org/doc/tutorials/objdetect/table_of_content_objdetect/table_of_content_objdetect.html) to see if there is something you can use.
White balance does not do anything to lighting, you should do adaptive thresholding or some kind of high pass filtering instead.
What I'm trying to do is have a background image, for sake of simplicity, lets say it's a picture of the front of a house. Then, I want to have a red ball move from window to window.
**I want to have a background picture, and a picture on top of it.
**I then want to be able to tell the top picture EXACTLY where to go.
How can I do this?
I'm just beginning to learn about animations in Android, and have not yet run across any way to do this.
There are two routes to animation in android: Canvas and OpenGL ES.
I would recommend OpenGL for anything requiring smoothness and speed, like a moving ball.
You should create a view using the helper class GLSurfaceView
http://android-developers.blogspot.com/2009/04/introducing-glsurfaceview.html, and implement a Renderer.
I assume you have the images saved in your res/drawable folders, in a format like png and the ball file contains an alpha channel.
You can see many tutorials online, but basically you need to load your background image and your ball resource at onSurfaceCreated and store it in a texture using GLUtils.texImage2D.
In the onDrawFrame method, you should set up a 2D projection such as glOrtho2D, then draw the background.
Then just before you draw the ball texture, you can use the glTranslate(x,y,0) function to move the ball over the house. Use an alpha blend for the ball:
glBlendFunc(GL_SRC_ALPHA, GL_SRC_ONE_MINUS_ALPHA);
glEnable(GL_BLEND);
Unfortunately writing in OpenGL isn't as straightforward as you might hope. Everything is done with 3D coordinates, despite the fact you want only a 2D image. But hopefully this gives you enough info to google for good exmaples, which are abundant!