Shulou(Shulou.com)06/01 Report--

Today, I would like to talk to you about what the characteristics of SUSAN refer to. Many people may not know much about it. In order to make you understand better, the editor has summarized the following for you. I hope you can get something from this article.

Today we will introduce another feature detection operator-SUSAN feature. SUSAN operator is a very nice name, in fact, SUSAN operator in addition to the name is good, she is also very practical, but also easy to use, SUSAN's full name is: Smallest Univalue Segment Assimilating Nucleus, on the translation of this noun domestic messy, such as the minimum kernel value similar region, the minimum same value shrinking kernel region and the minimum core value similar region and so on.

SUSAN operator is an efficient edge and corner detection operator, and has the noise reduction function of structure preservation.

Let's first take a look at the following picture:

The image is on a white background, an area with a deep color (dark area), and moves on the image with a circular template. if the gray value of the pixel in the template and the pixel in the center of the template (called kernel Nucleus) is less than a certain threshold, it is considered that the point has the same gray level as the kernel Nucleus, and the area composed of pixels that meet this condition is called USAN (Univalue Segment Assimilating Nucleus).

Next, let's analyze the USAN values of the five circular modules in the image above. For the e-circular template in the image above, it is completely in the white background, according to the previous definition of USAN, the USAN value at the template is the largest; with the movement of the template c and d, the USAN value decreases gradually; when the circular template moves to b, its center is on the edge line, and its USAN value is gradually reduced to half of the maximum; while the circular template runs to the corner a, the USAN value is the minimum. Therefore, from the above description: we can deduce that the USAN value of the point at the edge is less than or equal to half of the maximum value. From this, we can get the basic principle of SUSAN edge and corner extraction algorithm: the USAN value at the edge or corner is the smallest, and the location and direction information of the edge, corner and other features can be detected according to the size of the USAN region.

Let's use a formula to describe the previous content: the SUSAN operator moves across the image with a circular template whose radius is generally 37 pixels (3.4pixels). Each pixel in the template is compared with the central pixel as follows:

Where is the central pixel, is the other pixels within the mask, t is a pixel difference threshold (usually for areas with low contrast, select a smaller t; otherwise, t can choose a larger threshold). Then, the above formula is counted, and the statistical method is as follows:

The resulting n value is the size of the USAN. After the USAN value is obtained, the preliminary edge response can be obtained by thresholding. The formula is as follows:

Where g is:

That is, the value of g is the maximum value of 3 USAN 4. The lower the USAN value, the stronger the edge response.

When the initial edge response is suppressed by non-Maxima, the edge information of the image can be obtained. The previous image of SUSAN edge detection:

The above completed the function of edge detection by SUSAN.

To detect corners by using SUSAN operator:

Use the circular template to traverse the image and calculate the USAN value at each point

Set a threshold g, generally take the value as 1 Max (n), that is, take the value as half of the maximum value of USAN, threshold, and get the corner response

Use non-maximum suppression to find corners.

There are a lot of pseudo corners in the corners obtained in the above way. To remove pseudo corners, the SUSAN operator can be implemented in the following ways:

Calculate the center of gravity of the USAN area, and then calculate the distance between the center of gravity and the center of the template, which is not the correct corner if the distance is small.

Determine whether the pixels passed by the center of gravity of the USAN region and the connection in the center of the template belong to the pixels of the USAN region, and if so, then the point in the center of the template is the corner.

SUSAN operator is an operator with simple principle and easy to understand. Because its index is based on the gray comparison of the surrounding pixels, it does not involve the gradient operation at all, so its anti-noise ability is very strong, and the amount of computation is relatively small; at the same time, the SUSAN operator is also an isotropic operator; finally, by controlling the parameters t and g, it is easy to control the images with different contrast and different shapes by setting appropriate t and g according to the specific conditions. For example, if the contrast of the image is larger, a larger t value can be selected, while if the image contrast is smaller, a smaller t value can be selected. In a word, SUSAN operator is a very rare operator, which not only has a good performance of edge detection, but also has a good effect on corner detection.

After reading the above, do you have any further understanding of what SUSAN features refer to? If you want to know more knowledge or related content, please follow the industry information channel, thank you for your support.

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.