Impact of computer vision based secure image enrichment techniques on image classification model

Abstract

Growing advancement in Artificial Intelligence has led to developing systems which can think, analyze, recognize patterns and are capable enough to convert data into information has turned out to be a major prerequisite. An effort to provide enriched data with Computer Vision based Image Enrichment techniques and its impact on performance of Deep learning algorithms (VIRNet) is analyzed in the research. At first, various image enrichment technique like sharpening, denoising, color enhancement which makes the dominant features more visible which can be easily captured by the classification model. Later, popular computer vision techniques for image segmentation which differentiates the background and foreground area like Deeplabv3, Mask R-CNN and Region based segmentation methods are applied and its impact over Classification model is analyzed. The experimental study was conducted on popular image dataset Caltech 101 and Caltech-256, from the results it is clearly evident the image enrichment based on CLAHE outweighs the popular segmentation methods. Besides image segmentation and enrichment techniques, implementing an AES encryption in the process helps in protecting the sensitive data and ensures the data is kept private and secure. An image classification algorithm using deep learning equipped with AES encryption is implemented, which adds a layer of security to an image processing system.