Veterinary B-ultrasound Image Texture Feature Extraction

In view of the characteristics of veterinary B-ultrasound image texture, such as uneven light spots, strong and weak echoes, and uniform and uneven distribution, we intend to use computer image analysis technology to analyze and extract the texture features of veterinary B-ultrasound images.

The grayscale of veterinary B-ultrasound images reflects the size of the echo. The larger the grayscale value, the stronger the echo, and the smaller the grayscale value, the weaker the echo. Statistical methods are used to count the grayscale of veterinary B-ultrasound images, and the primary statistics such as grayscale mean, standard deviation, maximum value, minimum value, skewness, kurtosis and percentile are calculated. These statistics can basically reflect the grayscale distribution of veterinary B-ultrasound images. The grayscale mean can correspond to the strength of the echo. The larger the grayscale mean, the stronger the echo, and the smaller the grayscale mean, the weaker the echo. The skewness and kurtosis can reflect the characteristics of the grayscale distribution. The larger the skewness value, the more the grayscale distribution deviates from the grayscale mean, and the smaller the kurtosis, the more symmetrical the grayscale distribution. The larger the kurtosis, the more concentrated the grayscale distribution, and the smaller the kurtosis, the more dispersed the grayscale distribution.

The light spot is a small area in the veterinary B-ultrasound image that is bright in the middle and slightly dark around. The veterinary B-ultrasound image segmentation technology is used to segment the light spot area in the image. These light spot areas are called light spots. The size of the light spot can reflect the thickness of the light spot. The larger the size, the thicker the light spot, and the smaller the size, the thinner the light spot. There are many light spots in a veterinary B-ultrasound image. The size of a light spot cannot represent the thickness of the light spot in the veterinary B-ultrasound image. The average method is used to calculate the average value of all the light spot sizes in the veterinary B-ultrasound image, that is, the average size of the light spot, which can better reflect the thickness of the light spot in the image. The number of light spots can reflect the density of light spots. The larger the number of light spots, the denser the distribution of light spots in the veterinary B-ultrasound image, and the smaller the number of light spots, the sparser the distribution of light spots in the image.

Extraction of grayscale histogram of veterinary B-ultrasound images

Extraction of grayscale histogram of veterinary B-ultrasound machine. Most veterinary B-ultrasound machine images are grayscale images, so it is a natural choice to use grayscale features to retrieve veterinary B-ultrasound machine images. The grayscale histogram of a veterinary B-ultrasound machine image is a function that represents the frequency of each grayscale level in the veterinary B-ultrasound machine image. It is the number or relative value of pixels with this grayscale level in the veterinary B-ultrasound machine image. Compared with other features, grayscale features are very stable, insensitive to rotation, translation, and scale changes, and show strong robustness. At the same time, because it is easy to process in digital veterinary B-ultrasound machine images, it has become one of the very useful means. In the retrieval method, in addition to the color histogram, there are also color moments, color sets and other methods. There are currently two common methods: general histogram method and cumulative histogram method;

Shape feature extraction. Shape features are a significant feature of veterinary B-ultrasound image targets. Many queries may not target the color of veterinary B-ultrasound images, because the same object may have a variety of different colors, but its shape is always similar. An important criterion for shape feature representation is to require invariance to displacement, rotation, and scaling. A commonly used shape description method is the Fourier descriptor. Its main idea is to use the boundary after Fourier transformation as the shape feature. For the retrieval system of the general image library, it is difficult to use shape features for retrieval. So far, no exact mathematical description of the shape consistent with human vision has been found. For shape matching, Hough transform is the most classic method. In practical applications, due to the consideration of algorithm time and space complexity, shape expression and matching often use simpler methods, such as some shape parameter methods related to quantitative shape measurement (such as moment, area, perimeter, etc.).

link: https://www.bxlimage.com/nw/364.html