内窥镜图像质量在医生对早期病灶、异型增生复发的诊断中至关重要。因此本文根据血管对光谱的吸收特性, 提出了一种基于光谱变换的血管增强算法。首先, 该算法对图像RGB通道进行导向滤波, 将各通道分为亮度层和细节层; 接着, 将各通道的细节层进行基于信噪比的自适应增强, 并将亮度层进行拉伸, 使得GB通道的信息增强, R通道信息降低; 最后, 将各通道合并生成增强图像。本文应用该算法对大量内窥镜图像进行增强, 并且与Spectra B增强技术相比较。本文方法在DV-BV指标和韦伯对比度指标均优于Spectra B。

Endoscopic image quality plays an important role in the diagnosis of early lesions and dysplasia. Therefore, a blood vessel enhancement algorithm based on spectral absorption characteristics of blood vessels is proposed in this paper. First of all, RGB channels are obtained from the color image and divided into the brightness layer with the high dynamic range and the detail layer with the detail image information through the guided filter. Then, the detail layer of each channel is adaptively enhanced based on SNR (signal noise ratio), and the brightness layer is stretched to improve the GB channel information and to reduce R channel information. Finally each channel is merged to generate an enhanced image. In this article, a large number of endoscopic images is applied to this algorithm and compared with Karl Stroz's Spectra B enhancement technology. This method performs better in image enhancements while using the Detail Variance-Background Variance index and the Weber contrast index to evaluate.