Research on Image Compression Technique by Deep Learning Algorithm with Semantic Coding and Decoding

With the wide application of digital images in daily life and scientific fields, the transmission of these huge images faces blocks of limited bandwidth. Traditional image compression techniques tend to reduce the image size by distortion or loss of details while transmitting the zip file. However, the idea of semantic communication could be a new method to improve the image compression technique while expanding transmission capacity. This paper proposes a deep learning algorithm based on semantic coding and decoding, which can capture the content and meaning of the image more accurately and restore the details and textures of the original image, thus reducing compression-induced distortion. This paper utilizes a modified Convolutional Neural Network as the foundational model to extract semantic information from a large dataset of images and map it to transmission symbols in the practical transmission channel. While the Generative Adversarial Network is used to decode the coded representations and reconstruct the original images. The experimental results demonstrate that the proposed scheme exhibits higher robustness and superior compression performance when compared to JPEG and JPEG2000, which is evident in performance comparisons based on metrics like structural similarity index measurement and peak signal-to-noise ratio.