Superresolution Reconstruction Algorithm of Ultrasonic Logging Images Based on High-Frequency Enhancement

Table 1

Math symbols in the paper.

	Low-resolution image
	Original high-resolution image
	Feature extraction layer function
	Output of the first layer
	Function of multiscale dilated convolution block at stage
	Missing high-frequency information after the first stage training
	Transform of upsampling, downsampling, and converting to three-channel images
	Function of multiscale dilated convolution residual block at stage
	High-level feature of high-frequency information at stage
	High-level feature obtained at stage
	Convolution kernel size after expansion,
H	Function of upsampling and generating a three-channel image
	Output of the layer
	Convolutional layer function of the convolution kernel with expansion rate ,
	Convolutional layer function of the convolution kernel
	Output of feature fusion
	Output of the multiscale dilated convolution residual block
BN	Batch normalization
PReLU	Parametric rectified linear unit