Mirror reflection detection method based on color space and support vector machine

Abstract

The invention puts forward a mirror reflection detection method based on color space and a support vector machine. The method comprises the following main contents of a RGB (Red, Green and Blue) detection method, an HSV (Hue, Saturation and Value) detection method, color space selection by a SVM (Support Vector Machine), and an image inpainting method. The method comprises the following processesthat: firstly, in a detection stage, simultaneously using HSV and RGB color space information to segment mirror reflection; then, training a nonlinear support vector machine for selecting a color space according to image statistical characteristics extracted from each channel of the color space; then, importing a cost function used for detecting the mirror reflection; and finally, through a properreplacement plaque selection and plaque removal effect, applying a smooth method based on an edge to gradually eliminate an edge artifact so as to realize image inpainting. By use of the method, theHSV and RGB color space information and the nonlinear support vector machine are used for detecting and removing a mirror reflection area in colonoscopy and endoscope images, and the quality of the image is greatly improved.

Description

technical field [0001] The invention relates to the field of image processing, in particular to a specular reflection detection method based on a color space and a support vector machine. Background technique [0002] Specular reflections are bright pixels or spots with desaturated colors that can appear in medical imaging. This type of reflection is a common cause of image degradation during colonoscopy and endoscopy. Physicians often prefer to remove these specular reflections in medical images. In the case of endoscopy or colonoscopy, reflections are often produced in the presence of moisture on the inner surface of the GI tract due to the flashlight carried by the imaging device. Detecting these reflections is a challenging task due to their changing shape and color in different images. To remove these specular reflections, an image inpainting method can be used, which fills these blobs with similar shape and texture to the texture of the image in that area. By detec...

AI Technical Summary

Problems solved by technology

[0004] For the problem of image quality degradation, the object of the present invention is to provide a specular reflection detection method based on color space and support vector machine, first use HSV and RGB color space information to segment specular reflection in the detection stage, and then train a nonlinear A support vector machine

Images