Acoustic image texture feature extraction method based on multi-scale blurring

Abstract

The invention discloses an acoustic image texture feature extraction method based on multi-scale blurring, and belongs to the technical field of a computer. The acoustic image texture feature extraction method comprises image preprocessing, image eigenvalue extraction and classifier classification and comprises the main implementation steps of: searching and selecting a ROI (Region Of Interest) ofan image; carrying out adaptive median filtering processing on the image so as to minimize speckle noise of the image and keep robustness for impulse noise and edge characteristics of the image; carrying out expansion and corrosion operations on the image so as to further remove the speckle noise and soften the image; improving a contrast ratio of the image by utilizing histogram equalization, sothat the image has the same pixel number at each gray level; carrying out a LBP (Local Binary Pattern) blurring operation on the image to obtain a FLBP map corresponding to the ROI; extracting different scales of features by adopting DT-CWT (Dual Tree Complex Wavelet Transform); and selecting a RF (Random Forest) as a classifier, and dividing contents in the ultrasonic image into positive examples and counterexamples. The acoustic image texture feature extraction method can efficiently and accurately classify the ultrasonic image and has wide development prospect.

Description

technical field [0001] The invention belongs to the field of computer technology, and in particular relates to a multi-scale fuzzy-based acoustic image texture feature extraction method combining computer technology, image processing, artificial intelligence (AI) and data mining. Background technique [0002] Ultrasonic images reflect the differences in acoustic parameters in the medium, and can obtain information different from optics, X-rays, and y-rays. Ultrasound has a good ability to distinguish soft tissues of the human body, and can obtain useful signals with a dynamic range of up to 120dB, which is conducive to identifying small lesions in biological tissues. Ultrasound images can obtain the desired images without dyeing when displaying living tissues. [0003] Ultrasound imaging is to use ultrasonic beams to scan the human body, and obtain images of internal organs by receiving and processing reflected signals. There are many kinds of ultrasonic instruments common...

AI Technical Summary

Problems solved by technology

[0004] However, in general, ultrasound images are of low resolution and the edges of the image content are blurred

Therefore, how to extract the best features for describing ultrasound images is a huge challenge, and they will highly affect the accuracy and training time of classification results in machine learning

[0005] Currently existing feature extraction methods such as local phase quantization LPQ (Local Phase Quantization) and complete discriminative local features CDLF (Completed Discriminative Local Features) are not as good as the features proposed in the present invention in terms of precision and recall for ultrasonic image classification Extraction Method

Images