Gastritis Cold Or Heat Image Research Based On Provincial Features

Gastritis is the unified name for gastric mucosal inflammation, which is common, frequently-occurring and can be divided into acute and chronic. And the diagnosis of gastritis mainly depends on stomach endoscope and biopsy of gastric mucous membrane. Herbalist doctor divides gastritis into stomach cold and heat in terms of the adjustments to different external environment of stomach. It has been a long time since stomach endoscope used as the important criterion in diagnosis of gastritis. But there are some negative factors such as strong subjective dependence and poor repeatability because it mainly relies on doctors'direct observation which is qualitative rather than quantitative. So it is important and practical to research the objectification, quantification and standardization of stomach endoscope.It is the current tendency to analyze endoscope image automatically with the computer imaging processing method. Such method has a lot of features such as automation, efficiency, objectivity, quantification and standardization. So it will play an important role in the extraction, automatic classification and aided diagnosis of endoscopes'features.Proposed region-based features used in image processing the image of endoscopic gastritis. The computer imaging processing method based on regional characteristics develops quickly in recent years. Endoscopic gastritis is characterized by the image of the main color and texture recognition rather than boundaries and statistics. So, compared to traditional statistical methods, modeling, spectral method, the image processing features based on regional characteristics have higher accuracy.The values of different color channel in endoscopic gastritis images is calculated by applying the partial dual pattern, after extracted separated regional characteristics, using mutual information theory and combination of domain knowledge to determine the weight coefficient Regional characteristics component. And then, input eigenvectors into support vector machine to training and recognition. Fit the different colors vector reference the R, G, B model of color theory and combination of domain knowledge.According to the results of the experiments, the computer imaging processing method based on provincial characteristics takes advantage to traditional image texture analysis methods on high computation and lower accuracy. At the same time, the recognition capability and the operating speed is much stronger and faster.