关键词: 图像压缩, 矢量量化, 分形压缩, 深度学习

Abstract: Image compression is a key technology in the fields of image processing and communications and has long been a research hotspot in academia. This paper systematically reviews the basic concepts and principles of image compression, distinguishing between lossless and lossy compression and introducing various encoding techniques. With regard to traditional compression methods, techniques based on the discrete cosine transform, discrete wavelet transform, vector quantization, and fractal compression are comprehensively analyzed, with discussions on their respective advantages, disadvantages, and applicable scenarios. Although these methods have played significant roles in the field of image compression, their limitations have gradually become apparent with further technological developments. In the context of deep learning-based image compression, this paper focuses on the application of convolutional neural networks, recurrent neural networks, generative adversarial networks, as well as the recently emerging Transformer and diffusion model approaches. These methods achieve more efficient compression and image reconstruction by automatically learning image features. In terms of performance evaluation, key metrics such as compression ratio, peak signal-to-noise ratio, and structural similarity index are analyzed, and this paper discusses both the application prospects and the challenges faced by image compression technologies in various fields. Finally, this paper outlines future development directions and research trends in image compression technology, suggesting that with the integration of deep learning and emerging technologies, intelligent image compression will become a crucial development direction in the future.

Key words: image compression, vector quantization, fractal compression, deep learning