Neural JPEG: End-to-End Image Compression Leveraging a Standard JPEG Encoder-Decoder

Abstract

Recent advances in deep learning have led to superhuman performance across a variety of applications. Recently, these methods have been successfully employed to improve the rate-distortion performance in the task of image compression. However, current methods either use additional post-processing blocks on the decoder end to improve compression or propose an end-to-end compression scheme based on heuristics. For the majority of these, the trained deep neural networks (DNNs) are not compatible with standard encoders and would be difficult to deply on personal computers and cellphones. In light of this, we propose a system that learns to improve the encoding performance by enhancing its internal neural representations on both the encoder and decoder ends, an approach we call Neural JPEG. We propose frequency domain pre-editing and post-editing methods to optimize the distribution of the DCT coefficients at both encoder and decoder ends in order to improve the standard compression (JPEG) method. Moreover, we design and integrate a scheme for jointly learning quantization tables within this hybrid neural compression framework.Experiments demonstrate that our approach successfully improves the rate-distortion performance over JPEG across various quality metrics, such as PSNR and MS-SSIM, and generate visually appealing images with better color retention quality.

Compression of images with a mathematical approach based on sine and cosine equations and vector quantization (VQ)

“…In these experiments, the codebook sizes are set to 8, 16, 32, 64, 128, 256, 512, and 1024. Tables (1) to (5) show the PSNR index for different Bitrates for pentagonal images [36]. Evaluations show that the proposed method has a higher PSNR in the reconstruction of compressed images.…”

“…In this version (ISCA), according to Equation ( 9), a search model with spiral behavior of sine and cosine functions is added to the algorithm. In this case, the spiral problem space, as in Figure (5), spirals the space around the optimal solution. Mathematical functions such as Equation ( 9) are used to model this type of search:…”

“…In this process, an address of the code word which is similar to the input image is formed by the encoder. Then, this address is used by the decoder to retrieve the image [5].…”