Access the full text.
Sign up today, get DeepDyve free for 14 days.
K Dabov, A Foi, V Katkovnik, K Egiazarian (2007)
Image denoising by sparse 3-D transform-domain collaborative filteringIEEE Trans Image Process, 16
Y LeCun, L Bottou, Y Bengio, P Haffner (1998)
Gradient-based learning applied to document recognitionProc IEEE, 86
Y Chen, T Pock (2017)
Trainable nonlinear reaction diffusion: A flexible framework for fast and effective image restorationIEEE Trans Pattern Anal Mach Intell, 39
Y Sun, GG Yen, Z Yi (2019)
IGD indicator-based evolutionary algorithm for many-objective optimization problemsIEEE Trans Evol Comput, 23
D Renza, E Martinez, A Arquero (2013)
A new approach to change detection in multispectral images by means of ERGAS indexIEEE Geosci Remote Sens Lett, 10
T Blickle (2000)
Tournament selectionEvol Comput, 1
K Zhang, W Zuo, Y Chen, D Meng, L Zhang (2017)
Beyond a Gaussian denoiser: residual learning of deep CNN for image denoisingIEEE Trans Image Process, 26
Y Sun, B Xue, M Zhang, GG Yen (2019)
Completely automated CNN architecture design based on blocksIEEE Trans Neural Netw Learn Syst, 31
S Hochreiter (1998)
The vanishing gradient problem during learning recurrent neural nets and problem solutionsInt J Uncertain Fuzziness Knowl Based Syst, 6
L Zhang, L Zhang, X Mou, D Zhang (2011)
FSIM: a feature similarity index for image quality assessmentIEEE Trans Image Process, 20
Y Zhou, GG Yen, Z Yi (2019)
Evolutionary compression of deep neural networks for biomedical image segmentationIEEE Trans Neural Netw Learn Syst, 31
SM Lim, ABM Sultan, MN Sulaiman, A Mustapha, KY Leong (2017)
Crossover and mutation operators of genetic algorithmsInt J Mach Learn Comput, 7
Z Wang, AC Bovik, HR Sheikh, EP Simoncelli (2004)
Image quality assessment: from error visibility to structural similarityIEEE Trans Image Process, 13
RS Sutton, AG Barto (1999)
Reinforcement learningJ Cogn Neurosci, 11
Convolutional neural networks (CNNs) have achieved effective results in image denoising tasks. However, CNN architectures for image denoising tasks are mainly designed manually, which not only relies on CNN-related professional knowledge, but also requires adjustment to different datasets for competitive performance. Algorithms for automatically evolving CNN architectures have been proposed, but most of them are designed for solving image classification tasks and consume considerable computational time and resources. To address these issues, an efficient automatically evolving CNN architecture algorithm for image denoising tasks using genetic algorithm is proposed, which is called fast block-based evolutionary denoising CNN (FBE-DnCNN). In FBE-DnCNN, a genetic encoding strategy based on both deep and wide net blocks is designed to effectively represent the image denoising CNNs for automatic architecture design. With the purpose of solving time-consuming and resource-dependent problems, the partial dataset-based technology is used. A novel refined fitness evaluation method with prior knowledge on parameters of CNNs is designed to improve reliability. For better feature extraction of shallow network layers, convolutional operation, prevention of overfitting, and improvement of the representational capacity, the Feature Block, Transition Block, Dropout Block, and SENet module are introduced in FBE-DnCNN to generate problem-specific search space. With block-specific crossover and mutation, a local search near the good solution is implemented to find better solutions. Experiments show that FBE-DnCNN can evolve distinguished image denoising CNNs with deep and wide architectures in a very short time. FBE-DnCNN achieves competitive performance for the image denoising tasks with different noise levels compared to the traditional approaches, state-of-the-art CNN-based algorithms, and NAS-based methods.
Memetic Computing – Springer Journals
Published: Jun 1, 2023
Keywords: Image denoising; Convolutional neural network; Neural network architecture design; Evolving neural networks
Read and print from thousands of top scholarly journals.
Already have an account? Log in
Bookmark this article. You can see your Bookmarks on your DeepDyve Library.
To save an article, log in first, or sign up for a DeepDyve account if you don’t already have one.
Copy and paste the desired citation format or use the link below to download a file formatted for EndNote
Access the full text.
Sign up today, get DeepDyve free for 14 days.
All DeepDyve websites use cookies to improve your online experience. They were placed on your computer when you launched this website. You can change your cookie settings through your browser.