Hessian matrix, how to combine Ixx & Iyy together? - image-processing

"Before extracting the lines, you need to detect potential points on them. Apply a Gaussian filter first and use the Sobel filters as derivative operators. Threshold the determinant of the Hessian and then apply non-maximum suppression in 3 × 3 neighborhoods. Ignore pixels for which any of the filters falls even partially out of the image boundaries."
I understand to gaussian an image first to eliminate noise, then take twice with Sobel_x and Sobel_y, respectively, which became Ixx and Iyy in Hessian that would show horizontal line and vertical line in image.But how am I suppose to get Ixxyy? But how could I combine these two image together to make Ixxyy as the right bottom in Hessian matrix?

The two off-diagonal elements of the Hessian matrix are d^2/dxdy. That is, they are the first derivative along y applied to the first derivative along x.
If the top-left element is obtained by Sobel_x( Sobel_x( image )), and the bottom-right element is Sobel_y( Sobel_y( image )), then the two other elements are both Sobel_y( Sobel_x( image )) or, equivalently, Sobel_x( Sobel_y( image )) (note that these two should be identical).
Do take into account that negative values are important here, and you should thus be careful to compute the Sobel filter in a way that preserves those negative values—don't store them in an unsigned integer array!

Related

Confusion in first and second order derivatives in image processing

In image processing, the Laplacian filter adds the two second order derivatives, one in x direction and the other in y direction.
However, I am confused when we use first order derivative filters. In that case, we don't add the two first order derivatives. Instead we use the magnitude of the two first order derivatives, that is the L2 norm of the gradient.
I want to know why we don't add these two first order derivatives like Laplacian when we use first order derivative filters. Thanks a lot.
The Laplacian is defined as the trace of the Hessian matrix. The Hessian matrix collects all second-order derivatives, which include also things like d^2/dxdy. The diagonal of the Hessian are the second derivative along each axis. Thus, the trace is their sum. [You should look into the determinant of the Hessian, it’s an interesting operator too.]
The gradient is a vector, composed of the partial derivative along each axis. Its magnitude (norm) is the square root of the sum of the square elements.
These things are different because they have a different meaning and a different purpose.

Finding edges in a height map

I want to find sharp edges in a heightmap image, while ignoring shallow edges.
OpenCV offers multiple approaches to finding edges in a 2d Image: Canny, Sobel, etc.
However, all these approaches work by comparing the intensity values on both sides of the edge.
If the 2D Image represents a height map of a 3D object, then this results in some weird behaviour.
In a height map, the height of a 3D object at a given X/Y coordinate is represented as the intensity of the 2D Pixel at that X/Y coordinate:
In the above picture, at the edge B the intensity changes only slightly between the left and right side, even though it is a sharp corner.
At the edge A, there is a bigchange in the intensity between pixels on the left side of the edge and the right, even though it is only a shallow angle.
So there is no threshold for Canny or Sobel that will preserve the sharp edge but filter the shallow edge.
(In the above example, the edge B has one side with an ascending slope, and one side with a descending slope. I could filter for this feature; but that would remove the edges C and D as well)
How can I get a binary edge image, containing only edges above a certain angle? (e.g. edge B, C, and D, but not A)
Or alternatively, how can I get a gradient derivative image, where the intensity of each pixel is proportional to the angle of the edge at that pixel?
Probably you'll want to use second derivative instead of first for this task.
Here's my intuition: taking derivative of height (intensity in your case) at each position on an evenly spaced grid would be proportional to arctan of the surface slope between sampling points (or at sampling points if you use a 2-sided derivative approximation). But since you want to detect sharp edges - you are looking for a derivative of slope at the sampling points. This means that you can set a threshold on a derivative of arctan of derivative of intensity to achieve your goal (luckily there's no "need to go deeper" :) )
You will have to be extra careful with taking a derivative of "slope angles" that you'll get - depending on the coordinate system you may come across ambiguity of angle difference (there are 2 ways to get from one angle to another, which are different in general case; you're looking for the "shorter" one). You can look for possible solution here
I have a rather simple approach that I came across wile reading a blog post.
It involves computing the median value of the gray scale image. Using this value we can now set two threshold values:
lower: max(0, (1.0 - 0.33) * v)
upper: min(255, (1.0 + 0.33) * v)
Now pass these two values as parameters into the cv2.Canny() function.
You will now be able to perform an optimized edge detection given any image. The crux of this answer depends on the median value of the image which varies for different images.
If i understand your question correctly, "what you need is basically a corner with high intensity values".
If that is so then look for Harris corner detector which would help you to find points with high gradient change in both direction.
http://docs.opencv.org/2.4/doc/tutorials/features2d/trackingmotion/harris_detector/harris_detector.html
Once you detect the corners you can filter the corners which have high intensity by using a suitable threshold.

How to transform filter when using FFT to do 2d convolution?

I want to use FFT to accelerate 2D convolution. The filter is 15 x 15 and the image is 300 x 300. The filter's size is different with image so I can not doing dot product after FFT. So how to transform the filter before doing FFT so that its size can be matched with image?
I use the convention that N is kernel size.
Knowing the convolution is not defined (mathematically) on the edges (N//2 at each end of each dimension), you would loose N pixels in totals on each axis.
You need to make room for convolution : pad the image with enough "neutral values" so that the edge cases (junk values inserted there) disappear.
This would involve making your image a 307x307px image (with suitable padding values, see next paragraph), which after convolution gives back a 300x300 image.
Popular image processing libraries have this already embedded : when you ask for a convolution, you have extra arguments specifying the "mode".
Which values can we pad with ?
Stolen with no shame from Numpy's pad documentation
'constant' : Pads with a constant value.
'edge' : Pads with the edge values of array.
'linear_ramp' : Pads with the linear ramp between end_value and the arraydge value.
'maximum' :
Pads with the maximum value of all or part of the
vector along each axis.
'mean'
Pads with the mean value of all or part of the
vector along each axis.
'median'
Pads with the median value of all or part of the
vector along each axis.
'minimum'
Pads with the minimum value of all or part of the
vector along each axis.
'reflect'
Pads with the reflection of the vector mirrored on
the first and last values of the vector along each
axis.
'symmetric'
Pads with the reflection of the vector mirrored
along the edge of the array.
'wrap'
Pads with the wrap of the vector along the axis.
The first values are used to pad the end and the
end values are used to pad the beginning.
It's up to you, really, but the rule of thumb is "choose neutral values for the task at hand".
(For instance, padding with 0 when doing averaging makes little sense, because 0 is not neutral in an average of positive values)
it depends on the algorithm you use for the FFT, because most of them need to work with images of dyadic dimensions (power of 2).
Here is what you have to do:
Padding image: center your image into a bigger one with dyadic dimensions
Padding kernel: center you convolution kernel into an image with same dimensions as step 1.
FFT on the image from step 1
FFT on the kernel from step 2
Complex multiplication (Fourier space) of results from steps 3 and 4.
Inverse FFT on the resulting image on step 5
Unpadding on the resulting image from step 6
Put all 4 blocs into the right order.
If the algorithm you use does not need dyadic dimensions, then steps 1 is useless and 2 has to be a simple padding with the image dimensions.

How to apply box filter on integral image? (SURF)

Assuming that I have a grayscale (8-bit) image and assume that I have an integral image created from that same image.
Image resolution is 720x576. According to SURF algorithm, each octave is composed of 4 box filters, which are defined by the number of pixels on their side. The
first octave uses filters with 9x9, 15x15, 21x21 and 27x27 pixels. The
second octave uses filters with 15x15, 27x27, 39x39 and 51x51 pixels.The third octave uses filters with 27x27, 51x51, 75x75 and 99x99 pixels. If the image is sufficiently large and I guess 720x576 is big enough (right??!!), a fourth octave is added, 51x51, 99x99, 147x147 and 195x195. These
octaves partially overlap one another to improve the quality of the interpolated results.
// so, we have:
//
// 9x9 15x15 21x21 27x27
// 15x15 27x27 39x39 51x51
// 27x27 51x51 75x75 99x99
// 51x51 99x99 147x147 195x195
The questions are:What are the values in each of these filters? Should I hardcode these values, or should I calculate them? How exactly (numerically) to apply filters to the integral image?
Also, for calculating the Hessian determinant I found two approximations:
det(HessianApprox) = DxxDyy − (0.9Dxy)^2 anddet(HessianApprox) = DxxDyy − (0.81Dxy)^2Which one is correct?
(Dxx, Dyy, and Dxy are Gaussian second order derivatives).
I had to go back to the original paper to find the precise answers to your questions.
Some background first
SURF leverages a common Image Analysis approach for regions-of-interest detection that is called blob detection.
The typical approach for blob detection is a difference of Gaussians.
There are several reasons for this, the first one being to mimic what happens in the visual cortex of the human brains.
The drawback to difference of Gaussians (DoG) is the computation time that is too expensive to be applied to large image areas.
In order to bypass this issue, SURF takes a simple approach. A DoG is simply the computation of two Gaussian averages (or equivalently, apply a Gaussian blur) followed by taking their difference.
A quick-and-dirty approximation (not so dirty for small regions) is to approximate the Gaussian blur by a box blur.
A box blur is the average value of all the images values in a given rectangle. It can be computed efficiently via integral images.
Using integral images
Inside an integral image, each pixel value is the sum of all the pixels that were above it and on its left in the original image.
The top-left pixel value in the integral image is thus 0, and the bottom-rightmost pixel of the integral image has thus the sum of all the original pixels for value.
Then, you just need to remark that the box blur is equal to the sum of all the pixels inside a given rectangle (not originating in the top-lefmost pixel of the image) and apply the following simple geometric reasoning.
If you have a rectangle with corners ABCD (top left, top right, bottom left, bottom right), then the value of the box filter is given by:
boxFilter(ABCD) = A + D - B - C,
where A, B, C, D is a shortcut for IntegralImagePixelAt(A) (B, C, D respectively).
Integral images in SURF
SURF is not using box blurs of sizes 9x9, etc. directly.
What it uses instead is several orders of Gaussian derivatives, or Haar-like features.
Let's take an example. Suppose you are to compute the 9x9 filters output. This corresponds to a given sigma, hence a fixed scale/octave.
The sigma being fixed, you center your 9x9 window on the pixel of interest. Then, you compute the output of the 2nd order Gaussian derivative in each direction (horizontal, vertical, diagonal). The Fig. 1 in the paper gives you an illustration of the vertical and diagonal filters.
The Hessian determinant
There is a factor to take into account the scale differences. Let's believe the paper that the determinant is equal to:
Det = DxxDyy - (0.9 * Dxy)^2.
Finally, the determinant is given by: Det = DxxDyy - 0.81*Dxy^2.
Look at page 17 of this document
http://www.sci.utah.edu/~fletcher/CS7960/slides/Scott.pdf
If you made a code for normal Gaussian 2D convolution, just use the box filter as a Gaussian kernel and the input image will be the same original image not integral image. The results from this method will be same with the one you asked.

SIFT matches and recognition?

I am developing an application where I am using SIFT + RANSAC and Homography to find an object (OpenCV C++,Java). The problem I am facing is that where there are many outliers RANSAC performs poorly.
For this reasons I would like to try what the author of SIFT said to be pretty good: voting.
I have read that we should vote in a 4 dimension feature space, where the 4 dimensions are:
Location [x, y] (someone says Traslation)
Scale
Orientation
While with opencv is easy to get the match scale and orientation with:
cv::Keypoints.octave
cv::Keypoints.angle
I am having hard time to understand how I can calculate the location.
I have found an interesting slide where with only one match we are able to draw a bounding box:
But I don't get how I could draw that bounding box with just one match. Any help?
You are looking for the largest set of matched features that fit a geometric transformation from image 1 to image 2. In this case, it is the similarity transformation, which has 4 parameters: translation (dx, dy), scale change ds, and rotation d_theta.
Let's say you have matched to features: f1 from image 1 and f2 from image 2. Let (x1,y1) be the location of f1 in image 1, let s1 be its scale, and let theta1 be it's orientation. Similarly you have (x2,y2), s2, and theta2 for f2.
The translation between two features is (dx,dy) = (x2-x1, y2-y1).
The scale change between two features is ds = s2 / s1.
The rotation between two features is d_theta = theta2 - theta1.
So, dx, dy, ds, and d_theta are the dimensions of your Hough space. Each bin corresponds to a similarity transformation.
Once you have performed Hough voting, and found the maximum bin, that bin gives you a transformation from image 1 to image 2. One thing you can do is take the bounding box of image 1 and transform it using that transformation: apply the corresponding translation, rotation and scaling to the corners of the image. Typically, you pack the parameters into a transformation matrix, and use homogeneous coordinates. This will give you the bounding box in image 2 corresponding to the object you've detected.
When using the Hough transform, you create a signature storing the displacement vectors of every feature from the template centroid (either (w/2,h/2) or with the help of central moments).
E.g. for 10 SIFT features found on the template, their relative positions according to template's centroid is a vector<{a,b}>. Now, let's search for this object in a query image: every SIFT feature found in the query image, matched with one of template's 10, casts a vote to its corresponding centroid.
votemap(feature.x - a*, feature.y - b*)+=1 where a,b corresponds to this particular feature vector.
If some of those features cast successfully at the same point (clustering is essential), you have found an object instance.
Signature and voting are reverse procedures. Let's assume V=(-20,-10). So during searching in the novel image, when the two matches are found, we detect their orientation and size and cast a respective vote. E.g. for the right box centroid will be V'=(+20*0.5*cos(-10),+10*0.5*sin(-10)) away from the SIFT feature because it is in half size and rotated by -10 degrees.
To complete Dima's , one needs to add that the 4D Hough space is quantized into a (possibly small) number of 4D boxes, where each box corresponds to the simiéarity given by its center.
Then, for each possible similarity obtained via a tentative matching of features, add 1 into the corresponding box (or cell) in the 4D space. The output similarity is given by the cell with the more votes.
In order to computethe transform from 1 match, just use Dima's formulas in his answer. For several pairs of matches, you may need to use some least squares fit.
Finally, the transform can be applied with the function cv::warpPerspective(), where the third line of the perspective matrix is set to [0,0,1].

Resources