A Comparative Analysis of CNN Techniques on Application to Fundus Images

Abstract

In the medical images, grayscale values play a major role in revealing valuable information about the image's typical brightness. This is a key factor in differentiating the objectives efficiently and effectively. Even with a narrow range value. In the present work, grayscale as well as contrast value calculations are used as supplemental metrics for image characterization and visualisation. The contrast values are determined by standard deviation and quantify the variations in the pixel intensities. The additional metrics provide a comprehensive viewpoint on image characteristics, improving the comprehensibility of analysis based on CNN. This paper explores the complex functions of convolutional neural networks (CNNs) in image processing, with particular attention on three core layers: pooling, activation, and convolution. Gray scale and contrast values were statistically analysed and reported. The Pooling, Activation, Convolution techniques of CNN were compared for the image gray scale and contrast values.