Super-Resolution of Textual Images using Autoencoders for Text Identification

Pal, Subarno; Jana, Susovan; Parekh, Ranjan

doi:10.1109/aspcon.2018.8748679

Cited by 4 publications

(2 citation statements)

References 15 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Various machine learning architectures have been proposed, and they can be applied to the issue of image super-resolution [16][17][18][19][20][21]. Among them, the convolutional neural network (CNN) is regarded as an architecture that is most widely applicable and most efficiently suited.…”

Section: Deep Machine Learning Neural Network For Image Upscalingmentioning

confidence: 99%

Image Upscaling with Deep Machine Learning for Energy-Efficient Data Communications

2023

View full text Add to dashboard Cite

Advanced algorithms of image quality enhancement have been attracting substantial attention recently due to the successful business model of video streaming services. The extremely high image quality in video streaming demands a significant increase in the transmit data rate. In turn, the required ultrahigh data rate causes the saturation of the video streaming service network if there is no remedy for this situation. Compression algorithms have contributed to the energy-efficient transmission of data; however, they have almost reached the upper bound. The demand for ultrahigh image quality by the user is significantly increasing. Meanwhile, minimizing data transmission is inevitable in energy-efficient communications. Therefore, to improve energy efficiency, we propose to decrease the image resolution at the transmitter (Tx) and upscale the image at the receiver (Rx). However, standard upscaling does not yield ultrahigh-quality images. Deep machine learning contributes to image super-resolution techniques with the cost of enormous time and resources at the user end. Hence, it is inappropriate for real-time applications. With this motivation, this paper proposes a deep machine learning-based real-time image super-resolution with a residual neural network on the prevalent resources at the user end. The proposed scheme provides better quality than conventional image upscaling such as interpolation. The comprehensive simulation verifies that our scheme substantially outperforms the conventional methods, utilizing the seven-layer residual neural network.

show abstract

Section: Deep Machine Learning Neural Network For Image Upscalingmentioning

confidence: 99%

Image Upscaling with Deep Machine Learning for Energy-Efficient Data Communications

2023

View full text Add to dashboard Cite

show abstract

“…Hence this model is more suitable for all image processing applications to reconstruct a good quality image from its minimal representation. Pal et al (2018) built an auto-encoder-based SR model for text identification, The encoder in their model consists of two convolution layers with a filter kernel of size 3 × 3 is used and is then performed upsampling. The decoder is made up of convolution layers with a filter size of 2 × 2, with ReLU as the activation function for all except the last layer and sigmoid as the last layer's activation function.…”

Section: Auto-encoder Based Srmentioning

confidence: 99%

Image Super-Resolution using Auto-Encoders with Parallel Skip-Connections

Purushothaman¹,

Kadangote²

2022

Journal of Computer Science

View full text Add to dashboard Cite

Super-resolution images are highly desired when employed for numerous analytical purposes and obviously because of their superior attractive visual effect. To create a High-Resolution (HR) image from one or more Low-Resolution (LR) images, an image super-resolution technique is used. When dealing with natural environments and settings, it is not always easy to access high-resolution images. The main barriers to the same are limitations in acquisition methods. In the domains of forensics investigation, remote sensing, digital monitoring, and medical imaging high-resolution images are always required. Modern methods using deep learning models have improved performance when compared to classic image processing methods. Using a convolution auto-encoder architecture with six parallel skip connections, a novel technique for improving lower solution natural images of size 256 × 256 to high-resolution images is proposed in this paper. The use of parallel skip connections between the encoder and decoder allows the network to reconstruct high-resolution images while still extracting relevant features from low-resolution images. Furthermore, the network is skillfully tuned using the right filters and regularization techniques. To create high-resolution output images, the decoder component makes use of the features' reduced representation in latent space. The model was evaluated using DIV 2K, CARS DATA, Set5, Set14, and the General data set after being trained using various data sets like CARS DATA and DIV 2K. Peak Signal to Noise Ratio, Structural Similarity Index, Mean Squared Error, and model behavior on numerous data sets are used to compare the proposed method to various current methods. Results reveal that the suggested model works better than the existing method and is reliable.

show abstract