24-09-2016, 01:06 PM
1456060800-samplereport.docx (Size: 913.29 KB / Downloads: 6)
In this Harris Corner Detection is basically used for panorama stitching. This algorithm identify the corner points of the image having well defined position also we can say that point as interest point.
For detecting corners it basically calculate the difference in intensity for a displacement of(u,v) in all directions which is as follow :
E(u,v) = ∑_(x,y)▒〖w(x,y)[I(x+u,y+v)-I(x,y)]〗
Where, w(x,y) is window function and I(x,y) is intensity of pixels.
Now, we have to maximize the E(x,y) and this is done by using Taylor expansion for detection of corner points. After maximizing we will get :
E(u,v) = [u,v]M[■(u@v)]
Where, M = ∑_(x,y)▒█(w(x,y)[■(IxIx&IxIy@IxIy&IyIy)]@)
Ix and Iy are Image derivatives in x and y directions.
Sample images to show how Harris Corner Detection works –
Feature Point Correspondence
Speeded up Robust Features (SURF) is used to extract descriptors for frames identified around the corners, detected above.
This algorithm uses an approximation of the determinant of hessian blob detector. It can be calculated with the help of three operations using a precomputed integral image. This is several time faster than SIFT.
SURF algorithm is implemented in three main steps that are as follow –
1. Detection of Interest Point : Despite of using cascade filters in SIFT, SURF uses square shaped filters are used for approximation of Gaussian smoothing. For filtering Integral image is used which is defined as :
2. Local neighbourhood descriptor : Role of this descriptor is to provide unique and robust information of our image. And it is obtain when it describes the intensity distribution of the pixels with in the neighbourhood of the corner points.
3. Matching : After detection of interest points or corner points and comparing all the descriptor which are obtained from different images, we can find the matching pair easily.
Image Warping
After obtaining corner detecting images we have to make a panorama using these images. This is done by stitching the images together.