Efficient and Effective Context-Based Convolutional Entropy Modeling for Image Compression

Abstract

It has long been understood that precisely estimating the probabilistic structure of natural visual images is crucial for image compression. Despite remarkable success of recent end-toend optimized image compression, the latent code representation is assumed to be fully statistically factorized such that the entropy modeling is feasible. Here we describe context-based convolutional networks (CCNs) that exploit statistical redundancies in the codes for improved entropy modeling. We introduce a 3D zigzag coding order together with a 3D code dividing technique to define proper context and to achieve parallel entropy decoding, both of which boil down to place translation-invariant binary masks on convolution filters of CCNs. We demonstrate the power of CCNs for entropy modeling in both lossless and lossy image compression. For the former, we directly apply a CCN to binarized image planes for estimating the Bernoulli distribution of each code. For the latter, the categorical distribution of each code is represented by a discretized mixture of Gaussian distributions, whose parameters are estimated by three CCNs. We jointly optimize the CCN-based entropy model with analysis and synthesis transforms for rate-distortion performance. Experiments on two image datasets show that the proposed lossless and lossy image compression methods based on CCNs generally exhibit better compression performance than existing methods with a manageable computational complexity.

Image compression using singular value decomposition by extracting red, green, and blue channel colors

“…The quantization step will normalize the resulting matrix, taking the psychovisual properties into account [22], [23]. Next, proceed to the encoder process by simply converting the data from one format to another format or pixel-based format to get better image resolution [24], [25]. Lastly, the compressed image is reconstructed with the same resolution but difference data size.…”

Image Compression Based on Deep Learning: A Review

“…When the compression ratio hit 40 percent, reconstructed images could still be more than 99% identical in terms of MS-SSIM. M. Li et al [83]. Presents context-based convolutional networks (CCNs) to be precise and efficient.…”

Soft Compression for Lossless Image Coding Based on Shape Recognition

“…Recent works are mainly in the area of lossy compression, which are based on convolutional neural networks (CNNs) [ 25 , 26 , 27 , 28 , 29 ], recurrent neural networks (RNNs) [ 30 ], generative adversarial networks (GANs) [ 31 ] and the context model [ 32 , 33 , 34 ]. Learning-based lossless image compression methods [ 35 , 36 , 37 , 38 , 39 ] use neural networks instead of the traditional encoder and decoder to achieve image compression. PixelCNN [ 40 ] and PixelCNN++ [ 41 ] as well as the methods based on bits-back coding [ 42 , 43 ] and flow models [ 44 , 45 ] shorten the distance between information theory and machine learning.…”