Abstract:
Image compression plays a vital role in storing and transmitting digital media. In addition to traditional compression methods, there have been recent advancements in AI-based techniques. These methods are designed with specific objectives in mind, such as optimized image reconstruction or utilizing latent representations for computer vision tasks. In this study, we explore the variations among these AI-based codecs based on their objectives by tackling a classification problem. following that we focuses on creating an enhanced image compressor capable of performing three tasks: image compression, computer vision, and image processing. Specifically, we chose face recognition and resolution doubling as secondary tasks alongside image compression.
