Retina image enhancement method

Abstract

The invention discloses a retina image enhancement method. The retina image enhancement method comprises steps that dual tree complex wavelet decomposition for a retina image is carried out to acquire 6 high-frequency images and 2 low-frequency images; morphological operation of the low-frequency images acquired through decomposition is carried out to acquire enhanced low-frequency images; de-noising processing on detail portions of the high-frequency images acquired through decomposition is carried out to acquire high-frequency images after de-noising; dual tree complex wavelet inverse transformation processing on the enhanced low-frequency images and the high-frequency images after de-noising is carried out to acquire the retina image after enhancement. The method is advantaged in that the effect of improving the retina vascular image contrast is excellent, the visual effect is improved, and the excellent enhancement effect on renal vessels is realized.

Description

technical field [0001] The present invention relates to the field of image processing, in particular to a retinal image enhancement method. Background technique [0002] In medicine, retinal microvascular disease is a direct response to human diabetes, hypertension, cardiovascular disease, etc., and the acquisition of high-resolution retinal images has become particularly important. Usually, the obtained fundus retinal images have low pixel contrast, uneven local illumination, and complex and dense distribution of blood vessels, which are not conducive to doctors' diagnosis of patients' conditions. The fundus cameras commonly used in medicine today, due to the poor imaging conditions and the limitations of the inherent phase difference of the human eye, the quality of the retinal images obtained is not high. Noise, uneven contrast, unclear outlines, etc. hinder the acquisition of useful information. Before formally diagnosing the retinal optic disc, macula, blood vessels, ...

AI Technical Summary

Problems solved by technology

Rajamani et al. processed retinal images in the spatial domain through adaptive histogram equalization [5], and achieved good results in improving contrast, but the effect of removing background noise was not ideal.

Banerje et al. combined wavelet transform and bilateral filtering to process retinal images, the algorithm processing speed is faster, but the wavelet reconstruction of this method is not complete

The enhancement method proposed by Kumari et al. uses median filtering in the direction of blood vessels, but the scale selection is too single and not suitable for most images

Jia et al. applied the Hessian matrix to blood vessel detection, and used a linear filter function to process images, which is useful for blood vessel enhancement, but not conducive to the enhancement of surrounding lesions.

[0004] To sum up, most of the existing methods cannot improve the retinal image globally, and there are still defects in improving the visual quality of the image.

Images