Showing papers on "Channel (digital image) published in 2022"

Visibility enhancement of underwater images based on active polarized illumination and average filtering technology

Abstract: Underwater imaging is a crucial and challenging problem. The backscattering caused by particles in turbid water can severely degrade the image. In this paper, A novel image dehazing method is devised, which combines optics with image processing technology. This method can directly remove part of the backscattering through polarization technology, and then quickly solve the dehazing image through our algorithm. Experimental results show that our method is valid and robust for targets of diverse materials, grayscale images and color images in diverse scattering environments. Furthermore, our method takes only 2 percent of the time of the fast dark channel prior (DCP) method when processing 1 K resolution images, which meets the requirements of real-time.

A Novel DWT and Deep Learning Based Feature Extraction Technique for Plant Disease Identification

Abstract: Disease detection in plants has been proven to be a very cumbersome job due to numerous limitations for example noise, illumination variations, color variations etc. Therefore, robust feature extraction becomes a difficult task when colored images are utilized for classification. In this work, a novel feature extraction technique is proposed where large scale coefficients based on discrete wavelet transform (DWT) are extracted from each channel of digital colored leaf images. Then, further significant features are selected using principal component analysis (PCA) of selected approximation coefficients from red, green and blue components of plant leaf images using orthogonal wavelets. Deep neural networks (DNNs) are further utilized for analysis of robust DWT-PCA colored image features and classification purpose because of the advantages that they perform better with large datasets. DNN architectures are pre-trained architectures which are trained on ImageNet database which contains millions of real-life objects training features. In this work, six types of pre-trained DNN architectures are used for extensive experimental analysis. The proposed plant disease identification system for colored images yields an accuracy up to 99% and performed much better than the presently existing systems.

U-Net Medical Image Segmentation Based on Attention Mechanism Combination

Abstract: The accuracy of medical image segmentation is of great significance to the diagnosis of patients. With the development of deep learning, the segmentation of medical images using convolutional neural networks has become a research hotspot. After the U-Net model was proposed, it has gradually become a commonly used convolutional neural network model in the field of medical image segmentation. However, medical images have the characteristics of different shapes of target organs and the image boundaries are not easy to determine. These problems lead to poor segmentation performance of the U-Net model. In view of the above problems, the attention mechanism is introduced from the two dimensions of space and channel to improve the U-Net model. Use Dice coefficient and IOU (Intersection Over Union) as evaluation metrics to compare model performance on multiple medical image datasets. The experimental results show that the U-Net model after introducing the attention mechanism has a better segmentation effect.

Watermarking of Digital Image Based on Complex Number Theory

Abstract: The objective of this paperwork is to implement a watermarking algorithm using complex number theory. Here, each pixel of the original image is considered as the real part of each complex number, and the imaginary part corresponds to the pixel of the watermark. The matrix consists of the absolute value of the complex number treated as the watermarked image. The real and imaginary components of each complex number cannot be retrieved without the phase angle, which is treated as the key to a secured watermarked image. The phase angle of each complex number transmitted through a secured channel. At the receiving end, when both the magnitude and phase angle of each pixel estimated properly, then the real and imaginary components of each complex number corresponding to the pixel of original and watermarked images are determined with some accuracy.

A Proposal of Digital Image Steganography and Forensics Based on the Structure of File Storage

Abstract: Steganography is a technology for transmitting secret messages in public channel by embedding secret information into a digital carrier. It has a wide range of applications in the Internet security thus has attracted great interests. Among all kinds of digital carriers, digital image has outstanding performance. To the best of our knowledge, the researches on image steganography mainly focus on the use of the feature of digital image in the spatial or frequency domain, while there are few researches on hiding information through the file format structure. Taking JPEG image standard as an example, this paper analyzes the file format structure as well as the compression procedure and reports a novel proposal of image steganography and forensics. The proposal uses the redundant channel between any marker segments in the JPEG format structure to embed the secret information. The validity of the proposal has been proved by experiments, and further study has shown that this proposal has no limitation on the hiding capacity and has good practicability.

Adaptive color correction and detail restoration for underwater image enhancement

Abstract: Underwater images have different color casts due to different attenuation conditions, such as bluish, greenish, and yellowish. In addition, due to floating particles and special illumination, underwater images have problems such as the lack of detail and unnecessary noise. To handle the above problems, this paper proposes a new, to the best of our knowledge, three-step adaptive enhancement method. For the first step, adaptive color correction, the three channels are adjusted based on the intermediate color channel, which is calculated by considering the positional relationship of the histogram distribution. For the second step, denoise and restore details, we first transform the space to hue, saturation, value (HSV), a detailed restoration method based on the edge-preserving decomposition that restores the lost detail while removing the influence of some noise. For the third step, we improve the global contrast. Still in the HSV space, a simple linear stretch strategy is applied to the saturation channel. Experiments on the standard underwater image enhancement benchmark data set have proved that our method yields more natural colors and more valuable detailed information than several state-of-the-art methods. In addition, our method also improves the visibility of underwater images captured by low-light scenes and different hardware cameras.