NBI gastroscope image processing method based on key point detection

Abstract

Disclosed is an NBI gastroscope image processing method based on key point detection. Key points are extracted from preprocessed gastroscope images through an SIFT algorithm, then part of NBI gastroscope images are extracted as a training set, a feature bag is formed by all key points in the training set, and a vision dictionary is generated through a clustering algorithm. Key point classification statistics is carried out on a single NBI gastroscope image according to an actual dictionary as a global feature. Finally, a trained classifier is used for recognizing the global feature of the single NBI gastroscope image. Therefore, the NBI gastroscope image recognition process is completed, and due to the fact that the SIFT algorithm has the good robustness on the NBI gastroscope images with the complex imaging environment and gastric juice and food residue interference, the recognition accuracy of the NBI gastroscope images is increased, and calculation complexity is lowered.

Description

[0001] technology neighborhood [0002] The invention relates to image recognition, in particular to an NBI gastroscope image processing method based on key point detection. Background technique [0003] Computer recognition technology for gastroscopy images has been developed for many years. For example, in order to reduce the influence of complex environments, the image is divided into multiple blocks on average, and then local color features are extracted for classification and recognition. At present, in many studies, grayscale texture features are widely used, such as local binary patterns (LBP, Local Binary Patterns), texture spectrum and wavelet domain co-occurrence matrix (WDCMF, Wavelet Domain Co-occurrence Matrix Features, these methods play their own roles. The computer recognition system for gastroscope images is mostly based on the imaging analysis of gastroscope images. Generally, the contrast of gastroscope images is low, and the analysis difficulty is increase...

AI Technical Summary

Problems solved by technology

Most of the computer recognition systems for gastroscope images are based on the imaging analysis of gastroscope images. Generally, the contrast of gastroscope images is low, and the difficulty of analysis increases, resulting in slow recognition of gastroscope images and low accuracy.

Images