Hey everyone! I was wondering if anyone is aware of any programs I could download that could help me with image registration (or tips on what to look for while registering)!
Hey guys, I feel like I'm going crazy. For context I'm a wildlife biologist who is about as proficient at coding as you can get in my field (which isn't a whole compared to you guys, lol). I'm trying to do something that I would consider to be pretty dang trivial. I've got imagery from a FLIR T540 thermal camera that takes RGB and Thermal Imagery in tandem.
Here are the images I'll be referencing: https://imgur.com/a/jg1qXH5
My goal is to register/align (whatever the correct terminology is) the thermal imagery with the RGB imagery and then crop out all parts of the RGB imagery that isn't within the registered thermal image bounds.
This seems incredibly simple to me, since the camera is essentially a fixed capture set-up, so once I make the affine or homography matrix that successfully transforms one image pair, I can apply it to all future image pairs. But the ORB algo I've been using seems to make pretty bad matches between the two images (and yes, I do convert the RGB imagery to grayscale while matching).
But past just tinkering with the ORB algo, I feel like this shouldn't take any sort of automated feature matching. Is there something I can do with knowing the camera sensor resolutions and lens dimensions? I've tried just inputing user selected keypoints and then making a transformation, but to no avail. My registered output images (aside from being a little messed up, cut off a fair amount of the RGB image, which in my mind shouldn't really occur as the RGB image contains the entire scene and more that the thermal image has).
Does anyone have any ideas? Below is some example python code I've typed up/stack overflow copied. Any input is super helpful, I can't imagine this is really that difficult of a task. I'm just getting super stuck and maybe am not approaching it the correct way.
ORB function
MAX_FEATURES = 500
GOOD_MATCH_PERCENT = 0.15
def alignImages(im1, im2):
# Convert images to grayscale
im1Gray = cv2.cvtColor(im1, cv2.COLOR_BGR2GRAY)
im2Gray = cv2.cvtColor(im2, cv2.COLOR_BGR2GRAY)
# Detect ORB features and compute descriptors.
orb = cv2.ORB_create(MAX_FEATURES)
keypoints1, descriptors1 = orb.detectAndCompute(im1Gray, None)
keypoints2, descriptors2 = orb.detectAndCompute(im2Gray, None)
print("Printing keypoints1: ", keypoints1)
print("Printing keypoints2: ", keypoints2)
# Match features.
matcher = cv2.DescriptorMatcher_create(cv2.DESCHello guys! Check out this article I've just published explaining how deep deformable registration networks can help us perform one-shot medical image segmentation!
Feedback welcome!
https://www.sicara.ai/blog/one-shot-medical-image-segmentation
Hello,
I have multiple images of the night sky that I would like to stack. Has anyone here ever attempted something like this in R and how did you do it? Thank you.
Update: I am testing, but the displacement and imwarp functions from imager seem to be working to align the images.
Hello, I would like to share my take on the real-time image registration problem on GPU. Image registration is an important problem in computer vision and image processing. If we have two images of the same object, this problem can be formulated as: how can we determine a coordinate system transformation, that allows us to match the displayed object? In the post below I use the VkFFT library to accelerate the well-known FFT-based phase correlation method of translation, rotation and scaling detection to the point that matching can be performed in real-time. Feel free to ask any questions!
https://towardsdatascience.com/real-time-image-registration-on-gpu-with-vkfft-library-c4e47f8050a0
I may be a robot
Hello guys! Check out this article I've just published explaining how deep deformable registration networks can help us perform one-shot medical image segmentation!
Feedback welcome!
https://www.sicara.ai/blog/one-shot-medical-image-segmentation
Hello guys! Check out this article I've just published explaining how deep learning networks can help us perform deformable image segmentation!
Feedback welcome!
https://www.sicara.ai/blog/deformable-image-registration
Hello guys! Check out this article I've just published explaining how deep learning networks can help us perform deformable image segmentation!
Feedback welcome!
https://www.sicara.ai/blog/deformable-image-registration
