Learning-based fast tongue image detection and segmentation method for mobile phone applications

Abstract

The invention discloses a learning-based mobile phone application-oriented tongue image rapid detection and segmentation method, which includes the following steps: S1. Mobile phone terminal detection, which includes: S11. Establishing a mobile phone terminal Adaboost cascade classifier, S12. Extracting tongue image candidate regions , S13. Detect the image of the candidate area containing the tongue image; S2. Server-side detection, which includes: S21. Establishing a server-side Adaboost cascade classifier, S22. Image detection, S3. Server-side segmentation; It includes: S31. Superpixel segmentation , S32. Calculate the probability image, S33. Segment the probability image. The detection at the mobile phone side and the server side are both based on the learned Adaboost cascade classifier in the invention, which can accurately detect the tongue image in the image in real time, and accurately segment the tongue image by using superpixel segmentation suitable for mobile phone images.

Description

technical field [0001] The invention relates to the technical field of image processing, in particular to a learning-based rapid tongue image detection and segmentation method for mobile phone applications. Background technique [0002] Tongue image segmentation is an image processing technology applied to tongue diagnosis in traditional Chinese medicine. Traditional tongue image processing targets most of the image acquisition equipment set by the hospital itself. The default image already has a tongue image, and the light conditions are constant. The traditional tongue image Segmentation technology often uses segmentation based on the edge of the tongue image or segmentation based on the image color threshold. These tongue image segmentation methods have low accuracy for images from mobile phones. The reason is that mobile phones have limited computing speed and small computing memory. The light in the photo is not stable, and the image captured by the mobile phone is not ...

AI Technical Summary

Problems solved by technology

[0002] Tongue image segmentation is an image processing technology applied to tongue diagnosis in traditional Chinese medicine. Traditional tongue image processing targets most of the image acquisition equipment set by the hospital itself. The default image already has a tongue image, and the light conditions are constant. The traditional tongue image Segmentation technology often uses segmentation based on the edge of the tongue image or segmentation based on the image color threshold. These tongue image segmentation methods have low accuracy for images from mobile phones. The reason is that mobile phones have limited computing speed and small computing memory. The light in the photo is not stable, and the image captured by the mobile phone is not well illuminated, so the edges are not obvious enough

Images