Reversible autoencoder, codec method, and image compression method and device

Abstract

The present invention provides a reversible self-encoder, encoding and decoding method, and image compression method and device, wherein the reversible self-encoder includes: encoding signal separation module, cascade reversible encoding module, encoding signal synthesis module, decoding signal separation module, cascade The reversible decoding module and the decoding signal synthesis module, wherein: the signal separation module separates the input image and generates two signals, the cascaded reversible encoding module and the cascaded reversible decoding module process the two signals, and the signal synthesis module processes the processed The two signals are synthesized. The invention relates to colleagues of the reversible self-encoder, and also provides a method for applying it to image compression. Compared with the reference neural network, the method can reduce the amount of parameters and calculation by half under the premise of achieving the same compression effect.

Description

technical field [0001] The invention belongs to the field of digital image processing and image compression, and in particular relates to a reversible self-encoder based on a lifting structure, a reversible self-encoding method based on a lifting structure, an image compression method using the reversible self-encoding method, and an image compression device. Background technique [0002] In recent years, deep neural networks have shined in the field of image processing, and have achieved amazing results in a variety of image processing tasks. Among them, in the direction of image compression, some use deep neural networks to replace the transformation, quantization and entropy coding modules in traditional methods, and realize end-to-end image compression methods. This type of method performs well under the subjective evaluation index of the image. Under the condition of the same MS-SSIM value, the end-to-end image compression method can usually save 30%-50% of the code st...

AI Technical Summary

Problems solved by technology

[0003] The existing autoencoder (AutoEncoder) generally consists of two parts, the encoding unit and the decoding unit. Although the end-to-end method has brought about a huge performance improvement, this type of method usually contains a large number of neural network parameters that need to be trained, so it will Take up a lot of storage and computing resources

Images