Cited time in webofscience Cited time in scopus

Deep Block Transform for Autoencoders

Title
Deep Block Transform for Autoencoders
Author(s)
Jin, Kyong Hwan
Issued Date
2021-05
Citation
IEEE Signal Processing Letters, v.28, pp.1016 - 1019
Type
Article
Author Keywords
autoencoderBlock transformconvolutional neural networkimage representation
Keywords
Zero frequencyLearning systemsConvolutionSignal encodingAdjoint operatorsBlock transformsBlurry imagesConvolution kernelDictionary learningHigh resolutionSliding Window
ISSN
1070-9908
Abstract
We discover that a trainable convolution layer with a stride over 1 and kernel ≥ stride is identical to a trainable block transform. A block transform is performed when we use a convolution layer with a stride ≥ 2 and a kernel ≥ the stride. For instance, if we use the same widths, such as a 2 × 2 convolution kernel and stride-2, there are no overlaps between sliding windows, so this layer operates a block transform on the partitioned 2 × 2 blocks. A block transform reduces the computational complexity due to a stride ≥ 2. To keep the original size, we apply a transposed convolution (stride = kernel ≥ 2), an adjoint operator of a forward block transform. Based on this relationship, we propose a trainable multi-scale block transform for autoencoders. The proposed method has an encoder consisting of two sequential convolutions with stride-2, a 2× 2 kernel, and a decoder consisting of the encoder's two adjoint operators (transposed convolution). Clipping is used for nonlinear activations. Inspired by the zero-frequency element in the dictionary learning method, the proposed method uses DC values for residual learning. The proposed method shows high-resolution representations, whereas the stride-1 convolutional autoencoder with 3 × 3 kernels generates blurry images. © 1994-2012 IEEE.
URI
http://hdl.handle.net/20.500.11750/15400
DOI
10.1109/LSP.2021.3082031
Publisher
Institute of Electrical and Electronics Engineers
Files in This Item:

There are no files associated with this item.

Appears in Collections:
Department of Electrical Engineering and Computer Science Image Processing Laboratory 1. Journal Articles

qrcode

  • twitter
  • facebook
  • mendeley

Items in Repository are protected by copyright, with all rights reserved, unless otherwise indicated.

BROWSE