Zhe-Ming Lu’s research while affiliated with Zhejiang University and other places

To accurately extract tabular data, we propose a novel cell-based tabular data extraction model (TDEM). The key of TDEM is to utilize grayscale projection of row separation lines, coupled with table masks and column masks generated by the VGG-19 neural network, to segment each individual cell from the input image of the table. In this way, the text content of the table is extracted from a specific single cell, which greatly improves the accuracy of table recognition.

In this letter, we propose a single frame based method to remove the stripe noise, meanwhile preserving the vertical details. The key idea is to employ the side-window filter to performedge-preserving smoothing, and then accurately separate the stripe noise via a 1D column guided filter. Experimental results demonstrate the effectiveness and efficiency of our method.

This Letter focuses on deep learning-based monkeys' head swing counting problem. Nowadays, there are very few papers on monkey detection, and even fewer papers on monkeys' head swing counting. This research tries to fill in the gap and try to calculate the head swing frequency of monkeys through deep learning, where we further extend the traditional target detection algorithm. After analyzing object detection results, we localize the monkey's actions over a period. This Letter analyzes the task of counting monkeys' head swings, and proposes the standard that accurately describes a monkey's head swing. Under the guidance of this standard, the monkeys' head swing counting accuracy in 50 test videos reaches 94.23%.

Single image dehazing is a challenging ill-posed problem which estimates latent haze-free images from observed hazy images. Some existing deep learning based methods are devoted to improving the model performance via increasing the depth or width of convolution. The learning ability of Convolutional Neural Network (CNN) structure is still under-explored. In this paper, a Detail-Enhanced Attention Block (DEAB) consisting of Detail-Enhanced Convolution (DEConv) and Content-Guided Attention (CGA) is proposed to boost the feature learning for improving the dehazing performance. Specifically, the DEConv contains difference convolutions which can integrate prior information to complement the vanilla one and enhance the representation capacity. Then by using the re-parameterization technique, DEConv is equivalently converted into a vanilla convolution to reduce parameters and computational cost. By assigning the unique Spatial Importance Map (SIM) to every channel, CGA can attend more useful information encoded in features. In addition, a CGA-based mixup fusion scheme is presented to effectively fuse the features and aid the gradient flow. By combining above mentioned components, we propose our Detail-Enhanced Attention Network (DEA-Net) for recovering high-quality haze-free images. Extensive experimental results demonstrate the effectiveness of our DEA-Net, outperforming the state-of-the-art (SOTA) methods by boosting the PSNR index over 41 dB with only 3.653 M parameters. (The source code of our DEA-Net is available at https://github.com/cecret3350/DEA-Net .)

For covert communication in lossy channels, it is necessary to consider that the carrier of the hidden watermark will undergo multiple image-processing attacks. In order to ensure that secret information can be extracted without distortion from the watermarked images that have undergone attacks, in this paper, we design a novel fragmented secure communication system. The sender will fragment the secret data to be transmitted and redundantly hide it in a large number of multimodal carriers of messenger accounts on multiple social platforms. The receiver receives enough covert carriers, extracts each fragment, and concatenates the transmitted secret data. This article uses the image carrier as an example to fragment the text file intended for transmission and embeds it into a large number of images, with each fragment being redundant and embedded into multiple images. In this way, at the receiving end, only enough stego images need to be received to extract the information in each image, and then concatenate the final secret file. In order to resist various possible attacks during image transmission, we propose a strong robust image watermarking method. This method adopts a watermark layer based on DFT, which has high embedding and detection efficiency and good invisibility. Secondly, a watermark layer based on DCT is adopted, which can resist translation attacks, JPEG attacks, and other common attacks. Experiments have shown that our watermarking method is very fast; both the embedding time and the extraction time are less than 0.15 s for images not larger than 2000×2000. Our watermarking method has very good invisibility with 41dB PSNR on average. And our watermarking method is more robust than existing schemes and robust to nearly all kinds of attacks. Based on this strong robust image watermarking method, the scheme of fragmenting and hiding redundant transmission content into a large number of images is effective and practical. Our scheme can 100% restore the secret file completely under different RST or hybrid attacks, such as rotation by 1 degree and 5 degrees, scaling by 1.25 and 0.8, and cropping by 10% and 25%. Our scheme can successfully restore the secret file completely even if 30% of received images are lost. When 80% of received images are lost, our scheme can still restore 61.1% of the secret file. If all stego images can be obtained, the original text file can be completely restored.

In many construction sites, whether to wear the safety helmet directly affects the life safety of workers. Therefore, monitoring the wearing state of safety helmets has become an important auxiliary means of construction safety. However, most current safety helmet wearing state monitoring algorithms only distinguish workers who are wearing safety helmets from those who are not, which has high detection limitations and algorithm performance needs to be improved. In this paper, we innovatively apply fine-grained classification algorithms to classify the wearing state of safety helmets, and propose a progressive multi-granularity training strategy based safety helmet wearing state classification algorithm PMG-Helmet (Progressive Multi-granularity for Helmet, PMG-Helmet) for the six classification dataset of safety helmet wearing state. This algorithm achieves multi-granularity classification of helmet wearing state through a puzzle generator and a progressive training strategy, and introduces the MC-Loss(Mutual Channel Loss) function designed specifically for fine-grained classification tasks to improve algorithm performance. In the algorithm inference stage, this paper normalized the weights of the outputs of each stage of the PMG-Helmet algorithm, resulting in better combination accuracy. The experimental results show that the accuracy of this algorithm on the six classification dataset is 93.36%. Specifically, in order to further investigate the effectiveness of the algorithm, this study conducted separate studies on the finer subcategories of "wearing the helmet correctly" and "wearing the helmet but not fastening the chin strap" during the experimental phase, achieving an accuracy of 90.11%.

With the rise of short video and the development of self-media, video has become a type of frequently used data, thus the protection of video data is particularly important. Digital watermarking is a very effective digital authentication and copyright protection technology that hides a piece of data called watermark into digital multimedia content. However, most existing video watermark algorithms fail to deal with the videos under geometric distortions, and they only focus on copyright protection, ignoring the task of tamper detection. This paper proposes a multipurpose video watermarking algorithm that is resistant to rotation and scaling attacks while providing both copyright protection and tamper detection. We design a DFT template embedding method to make the watermarking algorithm resistant to rotation and scaling attacks efficiently. This method can detect the rotation and scaling that the video has undergone during the watermark extraction process, and then extract the copyright watermark information based on geometric correction. In addition, we propose a robust watermark embedding method with tamper detection function, which embeds copyright watermark information with validation bits in the DCT coefficients of I frames. By embedding the copyright watermark block by block, the algorithm can not only improve the robustness, but also enable the system to detect tampering. Experiments show that the proposed algorithm has good robustness to various attacks such as compression, filtering, rotation, scaling, frame rate changing and other inter-frame operations, and it outperforms previous geometrically robust algorithm by an average of 12.8% under geometric attacks on the normalized correlation values. Also, we show that our algorithm is effective in tamper detection by localizing the distorted areas in the video frames.

Until today, digital image watermarking has not been large-scale used in the industry. The first reason is that the watermarking efficiency is low and the real-time performance cannot be satisfied. The second reason is that the watermarking scheme cannot cope with various attacks. To solve above problems, this paper presents a multi-domain based digital image watermarking scheme, where a fast DFT (Discrete Fourier Transform) based watermarking method is proposed for synchronization correction and an IWT-DCT (Integer Wavelet Transform-Discrete Cosine Transform) based watermarking method is proposed for information embedding. The proposed scheme has high efficiency during embedding and extraction. Compared with five existing schemes, the robustness of our scheme is very strong and our scheme can cope with many common attacks and compound attacks, and thus can be used in wide application scenarios.