图像处理算法在Android智能手机上的性能

Question

I am looking forward to start a project that will use OCR , Object tracking and other Image processing algorithms on Android and I want to accelerate these algorithms using external hardware accelerators on FPGA using the Open Accessory API.

Do Image processing apps perform bad and needs custom hardware for acceleration ? Is there a resource to know about performance of image processing algorithms on smart phones and embedded systems without writing one ?

Answer 1

It's only viable to do optimization after you have done measurements for your particular case.

If you need HW acceleration you might want to check out renderscript . It gives you access to GPU hardware to perform generic computations.

Answer 2

JavaCV offers a number of image processing algorithms. It is basically a Java wrapper for OpenCV. I found this post regarding OpenCV and OCR: https://stackoverflow.com/questions/1284214/simple-ocr-programming-tutorials-articles . Performance really depends on the size of the image and processor on the device. Not sure about using a FPGA. Have you considered using the "cloud" to offload processing?

Answer 3

You can use OpenCV for image processing in Android. The best tutorial I can find is OpenCV setup on Eclipse . However if you do image processing in Java the results will be slow, so use second part of the tutorial to program in C then using JNI make the code run for android. But still a lot of image re sizing and defining region of interest is needed to make the program run in real time if you do object recognition.