4 ms·
I wish there was better software supported robotics cameras. You can buy small csi-mipi modules to capture frames from a ton of imaging sensors. But all the n
by mips_avatar 8y ago
I wish there was better software supported robotics cameras. You can buy small csi-mipi modules to capture frames from a ton of imaging sensors. But all the nice tricks Apple/Google use to improve image quality like capturing multiple underexposed images to create a clearer composite are not implemented in OpenCV. All these machine vision sdks are stuck to idea that you will only get a single matrix of pixels to glean info from, forgetting that you need to integrate pixel accuracy into your machine vision code!
- rtkwe 8y agoI think it's a question of division of responsibilities OpenCV focuses on the CV part of the equation. You don't have to feed OpenCV the raw camera feed there can be a separate step that does the kind of image processing you're talking about.
- mips_avatar 8y agoI usually use openCV capture. Which is nice because it works pretty well. Adding some capture modes like the one I listed would be nice, but I see how a separate capture software module might be better. edit: The CV part isn't really separate, how you capture affects the chance that the pixel is incorrect. This information is useful, maybe you'd want to integrate this into your CV.