Binary Wavelet Transform-Based Financial Text Image Authentication Algorithm
Abstract
This paper presents a novel financial text image security authentication algorithm based on binary wavelet transform. Initially, the algorithm preprocesses the original financial text image using binary wavelet transform. Subsequently, it categorizes the pixels in each low-frequency subband image block based on their flippability. The nonflippable pixels are then hashed to generate watermark information, which replaces the flappable pixels in the corresponding mapping block to embed the watermark into the low-frequency subband. By studying the characteristics of high-frequency coefficients, the algorithm identifies the flippable coefficients in the frequency domain and uses block mapping to embed high-frequency subband encryption watermark information. The authenticity of the image block is verified by comparing the consistency between the reconstructed watermark and the extracted watermark. Furthermore, the low-frequency-high-frequency subband joint judgment criterion is employed to enhance tampering detection performance. Our experimental results indicate that the structural similarity (SSIM) of financial text images embedded with watermarks using this algorithm is above 0.99, satisfying the human visual system's image quality requirements. Under different tampering attack rates, the missed-detection probability and false-detection probability of this algorithm are lower than those of existing methods. It is also compatible with financial text image compression algorithms, making it suitable for tampering detection of important financial text images.
