I need to be able to execute my own code (algorithms) on the video stream created by the deviceβs camera, for this I considered the capture of the OpenCV camera, unfortunately, it does not currently support Android 4.0.3 @Samsung G2, which is my target device, as alternatives I consider capture using a Java camera object and, using JNI, ~ Marshal ~ captured data in the native domain, this, however, imposes overhead marshaling (copying data), to avoid this, I considered rendering the captured image (preview) in GL Texture (using, for example, Camera.setPreviewTexture ), and by directly accessing GL Texture @βs native domain and avoiding unnecessary copying.
Is there a way to directly access TextureSurface @ native domain?
Any help would be appreciated.
Nadav In Sofina
source
share