问题
I'm developing an iPhone app to recognize some well known symbols from pictures. I'm basically following these tutorials http://aishack.in/tutorials/sudoku-grabber-with-opencv-detection/ and http://sudokugrab.blogspot.it/2009/07/how-does-it-all-work.html, using OpenCv for template matching and GPUImage for image processing.
When all images are with the same luminance level, I can adjust the threshold of GPUImageLuminanceThresholdFilter and all works smooth, but, of course, I can't be sure of the luminance.
So, I need a simple adaptive threshold filter, like the one in those tutorials, which calculate the luminance into the area surrounding each pixel.
The GPUImageAdaptiveThresholdFilter doesn't fit my needs, because it detects and sharps the edges, while I need to enhance the symbols.
How can I implement that kind of filter?
回答1:
Asked to, the awesome Brad Larson added a blur size property to the box blur, and modified the adaptive threshold filter, so it works as expected! Thanx @BradLarson!
来源：https://stackoverflow.com/questions/10688672/implementing-a-simple-adaptive-threshold-in-gpuimage