Compressing Sign Information in DCT-based Image Coding via Deep Sign Retrieval

TL;DR

The paper tackles the challenge of compressing the sign information of DCT coefficients in image coding, where signs are equiprobable. It proposes sign retrieval, inspired by phase retrieval, which recovers signs at the decoder from magnitude information and uses a DNN prior plus a projection onto convex sets to iteratively refine the signs, with a sign residual transmitted for exact recovery. Integrated into a JPEG-like pipeline, the method separates magnitudes and signs, restores signs from magnitudes, and transmits a residual; experiments show improved sign bitrate and lower computation cost compared to prior approaches. The work suggests extending sign retrieval to video coding and other transforms, highlighting a promising direction that blends phase retrieval concepts with deep learning for image and video compression.

Abstract

Compressing the sign information of discrete cosine transform (DCT) coefficients is an intractable problem in image coding schemes due to the equiprobable characteristics of the signs. To overcome this difficulty, we propose an efficient compression method for the sign information called "sign retrieval." This method is inspired by phase retrieval, which is a classical signal restoration problem of finding the phase information of discrete Fourier transform coefficients from their magnitudes. The sign information of all DCT coefficients is excluded from a bitstream at the encoder and is complemented at the decoder through our sign retrieval method. We show through experiments that our method outperforms previous ones in terms of the bit amount for the signs and computation cost. Our method, implemented in Python language, is available from https://github.com/ctsutake/dsr.

Figures (19)

• Figure 1: Our encoder and decoder.
• Figure 2: Sign retrieval.
• Figure 3: Circle for \ref{['eq:pr_dft_0']}
• Figure 4: Disk for \ref{['eq:pr_dft_1']}
• Figure 6: Two points for \ref{['eq:pr_dct_0']}