Medical Image Enhancement Using Super

Resolution Methods

Koki Yamashita and Konstantin Markov(B)

University of Aizu, Aizuwakamatsu, Fukushima 965-8580, Japan

{m5231120,markov}@u-aizu.ac.jp

Abstract. Deep Learning image processing methods are gradually gain-

ing popularity in a number of areas including medical imaging. Classi-
fication, segmentation, and denoising of images are some of the most
demanded tasks. In this study, we aim at enhancing optic nerve head
images obtained by Optical Coherence Tomography (OCT). However,
instead of directly applying noise reduction techniques, we use multi-
ple state-of-the-art image Super-Resolution (SR) methods. In SR, the
low-resolution (LR) image is upsampled to match the size of the high-
resolution (HR) image. With respect to image enhancement, the upsam-
pled LR image can be considered as low quality, noisy image, and the HR
image would be the desired enhanced version of it. We experimented with
several image SR architectures, such as super-resolution Convolutional
Neural Network (SRCNN), very deep Convolutional Network (VDSR),
deeply recursive Convolutional Network (DRCN), and enhanced super-
resolution Generative Adversarial Network (ESRGAN). Quantitatively,
in terms of peak signal-to-noise ratio (PSNR) and structural similarity
index (SSIM), the SRCNN, VDSR, and DRCN significantly improved
the test images. Although the ERSGAN showed the worst PSNR and
SSIM, qualitatively, it was the best one.

Keywords: Medical image processing · OCT image enhancement ·

Image super resolution

1 Introduction
In recent years, Deep Neural Networks (DNN) have shown great success in image
processing and analysis, outperforming humans in some tasks such as image clas-
sification [20]. It has been a matter of time, when DNNs would find their way
in the area of medical image processing. The enhancement of medical images is
a task of high practical value since many of the current MRI or CT images are
of low quality. Classical image enhancement methods are mostly based on his-
togram equalization techniques [19] which don’t work well with medical images.
Lately, there have been some studies where the DNN are used for image enhance-
ment [15] and MRI scans denoising [8].
In this work, we focus on enhancing or rather denoising images obtained by
Optical Coherence Tomography (OCT) [21]. The OCT technology has become
c Springer Nature Switzerland AG 2020
V. V. Krzhizhanovskaya et al. (Eds.): ICCS 2020, LNCS 12141, pp. 496–508, 2020.
https://doi.org/10.1007/978-3-030-50426-7_37
 Medical Image Enhancement Using Super Resolution Methods 497

a widely used tool for assessing optic nerve head tissues and monitoring many
ocular pathologies. However, the quality of OCT scans is hampered by mainly
speckle noise [7] as well as some other artifacts [1]. There exist some methods,
both hardware and software based, to denoise OCT scans. For example, the
multi-frame averaging [10] is a hardware technique which greatly improves the
image quality, but requires long scanning time. This inflicts discomfort and strain
in many patients. Software based image denoising approaches include filtering
[16] or some numerical methods [6].
So far, with respect to the OCT image processing, the usage of deep learning
has been limited to image segmentation [22] and classification [14]. The only
other work on OCT denoising we are aware of is [4].
The goal of the OCT image enhancement task is to improve the quality of
a single OCT scan to match the quality of multi-frame averaged image pro-
duced by the OCT device. This would greatly reduce the time needed to obtain
high-quality image, because one multi-frame scan can takes about 3 min while
a single scan - only few seconds. From machine learning point of view, this is a
supervised multiple regression task as depicted in Fig. 1, where the input is the
low quality (LQ) single scan and the output is an enhanced high quality (HQ)
image resembling the multi-frame OCT scan.

Fig. 1. The task of OCT scan enhancement. Low quality single scans are processed to
obtain high quality images resembling the multi-frame scans as closely as possible.

In [4], researchers try to solve this task by adding Gaussian noise to the HQ
multi-frame scans and use them as input to their denoising network based on
the popular U-net [17]. This approach avoids problems with the image regis-
tration, because often there is a misalignment between single scans and their
multi-frame counterparts. However, it ignores the actual speckle noise distribu-
tion which could be far from Gaussian and is OCT device dependent as well. Our
approach differs in two main ways. First, we don’t add artificial noise to the HQ
multi-frame scans, but use the original LQ single scans. This apparently requires
image registration which we performed using the excellent SimpleITK toolkit [2].
Second, we don’t use DNN architectures targeted at image denoising, but adapt
several state-of-the-art single images super resolution (SR) networks for the pur-
poses of our task. They include super-resolution Convolutional Neural Network
498 K. Yamashita and K. Markov

(SRCNN), very deep Convolutional Network (VDSR), deeply recursive Convolu-

tional Network (DRCN), and enhanced super-resolution Generative Adversarial
Network (ESRGAN). The way we use the SR networks for image enhancement
and some details for each of them are given in the next section. Later, we describe
our data, experimental conditions and results we obtained.

2 Single Image Super Resolution

Single image super resolution (SR) is a classical problem in computer vision

where the aim is to recover high-resolution (HR) image from a single low-
resolution (LR) image. With the rise of deep convolutional networks, the num-
ber of proposed solutions and network architectures has increased dramatically
[24,26]. In practice, since the HR image size is bigger, during processing, the
input LR image has to be upsampled to match the size. There are different
strategies where and how to do this in the processing pipeline. Two widely used
approaches are shown in Fig. 2. In the first one, the LR image is upsampled in
advance using some form of interpolation and then is passed to the SR model as
in Fig. 2(a). The other way is to keep the LR image size and perform upsampling
at the last processing step as in Fig. 2(b).

(a) Pre-upscaling SR (b) Post-upscaling SR

Fig. 2. Two widely used SR architectures where image upsampling is done either before
a) or after b) the processing.

Since in our task, the size of the image should not change, we cannot use
those SR architectures directly. However, if we remove the upsampling step in
the case of Fig. 2(a), we end up with a system that essentially enhances the input
image without changing its size. This is illustrated in Fig. 3(a). Unfortunately,
this approach does not work with the architecture of Fig. 2(b). In this case,
the upsampling step is part of the processing pipeline and its parameters are
trainable. We solve this problem by first downsampling the input image and
then passing it to the system as shown in Fig. 3(b).
In the next four subsections we describe briefly each of the SR networks we
used in this study.
 Medical Image Enhancement Using Super Resolution Methods 499

(a) In pre-upscaling SR, the first up- (b) In post-upscaling SR, a new down-
sampling block is deleted sampling block is added

Fig. 3. Changes made to accommodate the two SR architectures for image enhance-
ment purposes.

2.1 Super Resolution Convolutional Neural Network (SRCNN)

The SRCNN [5] is a simple network consisting of two hidden convolutional layers
as can be seen in Fig. 4. The input is supposed to be the upscaled version of
the LR image, so the architecture corresponds to the pre-upsampling SR from
Fig. 2(a).

Fig. 4. SRCNN architecture.

Each hidden layer performs standard convolutional operation with output

clipped to be positive. The loss function is the mean squared error (MSE)
between the output image Ỹi and the target HR image Yi averaged over the
training set:
1
n
L(Θ) =  Yi − Ŷi 2 (1)
n i=1
The MSE loss function favors a high peak signal-to-noise ratio (PSNR) which
is a widely-used metric for quantitative evaluation of SR quality. However, the
PSNR is only partially related to the perceptual quality and in practice, some-
times images with high PSNR don’t look perceptually very good.
500 K. Yamashita and K. Markov

2.2 Very Deep Convolutional Network (VDSR)

Based on the popular VGG network [18] for image classification, the VDSR [11]
consists of many convolutional layers with ReLU activation. The residual connec-
tion between the input and the last hidden layer (the long line in Fig. 5), forces
the network to learn only the difference between the input and the target and
as a result allows network to be much more deeper without vanishing/exploding
gradients problem.

Fig. 5. VDSR architecture.

The input is an upsampled interpolated low-resolution (ILR) image, so the

VDSR architecture falls into the pre-upsampling SR category as in Fig. 2(a). The
loss function is computed as the Euclidean distance between the reconstructed
image and the HR target image similar to Eq. (1). Therefore, the VDSR as the
SRCNN favors high PSNR, but not high perceptual quality.

2.3 Deeply Recursive Convolutional Network (DRCN)

The VDSR [12] makes use of the same convolutional block up to 16 times. The
main difference from the other structures is that a multi-supervised strategy is
applied, so that the outputs of all the blocks are combined together as shown
in Fig. 6. This approach not only allows gradients to flow easily through the
network, but also encourages all the intermediate representations to reconstruct
the HR image. In such multi-supervised approach, there are multiple objectives
to minimize. The loss for the intermediate outputs is defined as:

1 
D N
l1 (θ) =  yi − ŷid 2 (2)
2DN i=1
d=1

where D is the number of recursions. For the final output with is a weighted
sum of all intermediate outputs the loss is:

1  
N D
l2 (θ) =  yi − wd ŷid 2 (3)
2N i=1
d=1
 Medical Image Enhancement Using Super Resolution Methods 501

Fig. 6. DRCN architecture.

The final loss function includes both the l1 and l2 as well as a regularization
term:
L(θ) = αl1 (θ) + (1 − α)l2 (θ) + β  θ 2 (4)
where α controls the trade-off between the intermediate and final losses and β -
the amount of regularization. Note that all losses use the MSE criterion, so the
DRCN also favors high PSNR images.

2.4 Enhanced Super Resolution Generative Adversarial Network

(ESRGAN)
The ESRGAN [23] is an improved version of the super resolution generative
adversarial network (SRGAN) [13]. It consists of two networks - Generator and
Discriminator working together. The structure of each of them is shown in Fig. 7.
The Generator includes multiple blocks called residual in residual dense block
(RRDB) which combine multi-level residual network and dense connections. The
upsampling block is located at the end of the pipeline, so the ESRGAN architec-
ture is of the type shown in Fig. 2(b). The Discriminator has a simpler structure

Fig. 7. ESRGAN architecture.

502 K. Yamashita and K. Markov

consisting of multiple convolution layers each followed by a batch normalization

and Leaky ReLU activation. One important difference between the ESRGAN and
other SR networks described above is that the Generator utilizes an improved
version of the so called perceptual loss [9]. Originally, it is defined on the acti-
vation layers of a pre-trained network where the distance between two activated
features is minimized. Thus, the Generator total loss is expressed as:

G = Lpercep + λLG + ηL1

Ltot (5)

where L1 = Ex  G(x) − y 1 is the 1-norm difference between the Generator

output G(x) given input image x and the target HR image y. Using such loss
makes the ESRGAN to produce images of higher perceptual quality than the
PSNR oriented networks.

3 Performance Evaluation

There exist various quantitative performance metrics adopted in image process-

ing among which the peak noise-to-signal ratio (PSNR) and the structural sim-
ilarity index measure (SSIM) [25] are the most widely used. In [4], authors used
pure SNR and SSID metrics, while we utilize the PSNR and SSID.
The MSE and PSNR between ground truth image I and reconstructed image
Iˆ both of which have N pixels as defined as:

1 
N
M SE = ˆ 2
(I(i) − I(i)) (6)
N i=1
L2
P SN R = 10 log( ) (7)
M SE
where L = 255 for 8-bit pixel encoding. Typical PSNR values vary from 20 to
40, higher is better.
On the other hand, the SSID is defined as:

ˆ = (2μI μIˆ + C1 )(2σI Iˆ + C2 )

SSIM (I, I) (8)
(μ2I+ μ2Iˆ + C1 )(σI2 + σI2ˆ + C2 )

where C1 = (k1 L)2 , C2 = (k2 L)2 are constants for avoiding instability, k1 
1, k2  1 are small constants, and μ and σ 2 are the mean and variance of the
pixels intensity.

4 Experiments
4.1 Database

For the experiments, we used a small database of about 350 OCT scans. Some
of the HQ multi-frame scans had several corresponding LQ single scans, so the
 Medical Image Enhancement Using Super Resolution Methods 503

same targets were used for those LQ images. Most of the HQ/LQ pairs required
alignment and for this purpose we used the SimpleITK image registration toolkit
[2]. Six HQ/LQ pairs were selected for testing, and the remaining data were split
into training and validation sets by 9:1 ratio.
Since the number of scans is quite small, we did exhaustive data augmentation
which includes horizontal and vertical flips, rotation by several different degrees,
etc., commonly used in image processing practice. In addition, each scan was
cropped into non-overlapping sub-images of size 224 × 224. Thus, we managed
to increase the number of training data roughly 100 fold.

4.2 Results

Here, we present the results in terms of PSNR and SSIM metrics for each of
the network architectures described in Sect. 2. In each case, we tried to tune
the network hyper-parameters to achieve the best possible result. The results
shown in the tables below reflect the performance dependence on the two most
impactful parameters we found for each network.
All the networks were trained with up to 100 epochs and for testing we used
the model obtained from the epoch where the PSNR of the validation data was
the highest.

SRCNN Results. The SRCNN is trained using small patches of size 33 × 33

taken from the input image with stride 14. This network in known to take many
training iterations to achieve good performance, so we chose a small learning
rate of 5.0e-6. We found that the batch size and the size of the filter of the first
convolutional layer have the biggest influence on the SRCNN performance. The
obtained PSNR and SSIM values are given in Table 1.

Table 1. SRCNN performance in terms of PSNR (dB) and SSIM.

Metric Batch size f1 size

7×7 9×9 11 × 11
64 25.23 25.15 24.81
PSNR 128 25.18 24.93 23.76
256 24.96 25.10 24.38
64 0.794 0.798 0.797
SSIM 128 0.795 0.798 0.790
256 0.791 0.796 0.792
504 K. Yamashita and K. Markov

VDSR Results. The patch size during the VDSR training was set to 41 × 41
with no overlap. We experimented with the number of convolutional blocks and
the batch size. The learning rate was set to 0.001 and the other hyper-parameters
were used as recommended by the VDSR developers. Table 2 shows the PSNR
and SSID values obtained during the experiment.

Table 2. VDSR performance in terms of PSNR (dB) and SSIM.

Metric Batch size Number of blocks

8 16 32
32 25.10 25.22 24.18
PSNR 64 25.12 25.38 25.30
128 24.18 25.45 25.51
32 0.791 0.785 0.543
SSIM 64 0.791 0.789 0.779
128 0.791 0.795 0.778

DRCN Results. With the DRCN, we used the same patch size as for the
VDSR, but with stride 21 [11]. Initially, the learning rate was set to 0.01 and dur-
ing training was decreased 10 times every time validation performance plateaus.
The main architectural hyper-parameters of the DRCN are the number of blocks
and the number of filters in each block. We varied those parameters and the
results with batch size of 128 are presented in Table 3.

Table 3. DRCN performance in terms of PSNR (dB) and SSIM.

Metric Filter number Number of blocks

4 8 16
16 24.81 25.02 25.12
PSNR 32 24.48 24.26 23.02
48 18.07 22.37 25.77
16 0.768 0.774 0.778
SSIM 32 0.761 0.762 0.535
48 0.723 0.535 0.687

We have to note that we could not find a good trade-off between the interme-
diate loss l1 and final loss l2 functions given in Eq. (2) and Eq. (3) respectively.
The best results we obtained when the combination parameter λ from Eq. (4)
was set to 0.
 Medical Image Enhancement Using Super Resolution Methods 505

ESRGAN Results. In terms of parameters, this is the biggest network among

all the networks we experimented with, and so is the number of possible hyper-
parameters. Structurally, for the generator, important are the RDDB number,
the RDB number in each RDDB as well as the number of convolutional layers
and the number of filters. The discriminator’s structure has no big influence on
the performance. As can be seen from Table 4, in our case, the RDDB number
and the filter number were the most sensitive to the ESRGAN performance. We
couldn’t obtain results for the case of RDDB number = 7 and filter number =
16 since the model was so big and did not fit in our GPU memory. The other
parameters were as follows: number of RDBs inside each RRDB = 6, number of
convolutional layers inside a RDB = 4, learning rate = 4.0e-4 with decay factor
of 2. For training and evaluation of the ESRGAN, we used the ISR toolkit [3]
and all the other parameters we left at their default values.

Table 4. ESRGAN performance in terms of PSNR (dB) and SSIM.

Metric RDDB RDB filter number

number 4 8 16
3 19.56 19.01 18.98
PSNR 5 19.53 21.25 18.92
7 19.64 18.69 NA
3 0.670 0.639 0.725
SSIM 5 0.432 0.722 0.730
7 0.658 0.377 NA

Networks Comparison. Here, we compare the best obtained performance

from all the networks we evaluated in terms of PSNR and SSIM. Figure 8 shows
bar plots for each metric together with the case when no enhancement is applied.
In terms of PSNR, the DRCN achieved the best result, while the best SSIM was
achieved by the SRCNN and VDSR. In both cases, the obtained metrics values
are much better than the baseline, i.e. the case of unprocessed single scan images.

(a) Best PSNR in dB. (b) Best SSIM.

Fig. 8. Comparison of the networks best performances in terms of PSNR and SSIM
with the baseline (“No Enhan.”)
506 K. Yamashita and K. Markov

Fig. 9. Example test single scan (first row, left), the corresponding multi-frame aver-
ages scan (first row, center), and the results from each network.

The ERSGAN, however, showed PSNR even lower than the baseline. This
can be explained with the fact that the ESRGAN is trained to improve the
perceptual loss more than the mean absolute error (MAE) which is the L1 in
Eq. (5) and is related to the PSNR. To verify this hypothesis, we looked at all
the test images enhanced by each of the networks and visually compared them.
Indeed, the ESRGAN has produced the best looking images with sharper edges
and higher contrast. As an example, we show one of the test single scans and
its corresponding multi-frame scan as well as its enhanced versions by all the
networks in Fig. 9.

5 Conclusion

In this study, we focused on enhancing single scans obtained from Optical Coher-
ence tomography. They all contain speckle noise as well as some other artifacts
making the interpretation of the OCT data cumbersome. Many OCT devices
apply multi-frame averaging techniques to alleviate this problem, but this app-
roach requires a lot of time and causes great discomfort to the patients.
Instead of using enhancing/denoising methods directly, we adopted some of
the state-of-the-art deep neural networks designed for image super resolution.
Since in many cases the low resolution images are first upscaled, an operation
that degrades their quality, the SR networks essentially enhance those upscaled
low resolution images.
We experimented with several SR networks such as SRCNN, VDSR, DRCN
and ERSGAN and evaluated them quantitatively using PSNR and SSIM metrics.
Since all the networks but ESRGAN use MSE based loss function, they all
achieved high PSNR values. However, qualitatively, the ESRGAN produced the
best looking images which we attribute to the use of a perceptual loss function.
 Medical Image Enhancement Using Super Resolution Methods 507

Our results are still preliminary, because the amount of training data was
clearly insufficient to reliably train big networks such as DRCN or ESRGAN.
Also, the OCT scans come from healthy patients only and many pathological
artifacts haven not been learned. In addition, we expect scans from different
OCT devices to have different noise distributions. All these problems we intend
to address in the future.

Acknowledgment. We are grateful to Prof. Sekiryu from Fukushima Medical Uni-

versity for providing the OCT scan data.

References
1. Asrani, S., Essaid, L., Alder, B.D., Santiago-Turla, C.: Artifacts in spectral-domain
optical coherence tomography measurements in glaucoma. JAMA Ophthalmol.
132(4), 396–402 (2014)
2. Beare, R., Lowekamp, B., Yaniv, Z.: Image segmentation, registration and char-
acterization in R with simpleITK. J. Stat. Softw. 86(8), 1–35 (2018). https://doi.
org/10.18637/jss.v086.i08
3. Cardinale, F., John, Z., Tran, D.: ISR (2018). https://github.com/idealo/image-
super-resolution
4. Devalla, S.K., et al.: A deep learning approach to denoise optical coherence tomog-
raphy images of the optic nerve head. Sci. Rep. 9(1), 1–13 (2019)
5. Dong, C., Loy, C.C., He, K., Tang, X.: Image super-resolution using deep convo-
lutional networks. IEEE Trans. Pattern Anal. Mach. Intell. 38(2), 295–307 (2016)
6. Du, Y., Liu, G., Feng, G., Chen, Z.: Speckle reduction in optical coherence tomog-
raphy images based on wave atoms. J. Biomed. Opt. 19(5), 056009 (2014)
7. Esmaeili, M., Dehnavi, A.M., Rabbani, H., Hajizadeh, F.: Speckle noise reduction
in optical coherence tomography using two-dimensional curvelet-based dictionary
learning. J. Med. Signals Sensors 7(2), 86 (2017)
8. Jiang, D., Dou, W., Vosters, L., Xu, X., Sun, Y., Tan, T.: Denoising of 3D mag-
netic resonance images with multi-channel residual learning of convolutional neural
network. Japan. J. Radiol. 36(9), 566–574 (2018)
9. Johnson, J., Alahi, A., Fei-Fei, L.: Perceptual losses for real-time style transfer
and super-resolution. In: Leibe, B., Matas, J., Sebe, N., Welling, M. (eds.) ECCV
2016. LNCS, vol. 9906, pp. 694–711. Springer, Cham (2016). https://doi.org/10.
1007/978-3-319-46475-6 43
10. Kennedy, B.F., Hillman, T.R., Curatolo, A., Sampson, D.D.: Speckle reduction in
optical coherence tomography by strain compounding. Opt. Lett. 35(14), 2445–
2447 (2010)
11. Kim, J., Kwon Lee, J., Mu Lee, K.: Accurate image super-resolution using very
deep convolutional networks. In: Proceedings of the IEEE Conference on Computer
Vision and Pattern Recognition, pp. 1646–1654 (2016)
12. Kim, J., Kwon Lee, J., Mu Lee, K.: Deeply-recursive convolutional network for
image super-resolution. In: Proceedings of the IEEE Conference on Computer
Vision and Pattern Recognition, pp. 1637–1645 (2016)
13. Ledig, C., et al.: Photo-realistic single image super-resolution using a generative
adversarial network. In: Proceedings of the IEEE Conference on Computer Vision
and Pattern Recognition, pp. 4681–4690 (2017)
508 K. Yamashita and K. Markov

14. Lee, C.S., Baughman, D.M., Lee, A.Y.: Deep learning is effective for classifying
normal versus age-related macular degeneration OCT images. Ophthalmol. Retin.
1(4), 322–327 (2017)
15. Lu, L., Zheng, Y., Carneiro, G., Yang, L. (eds.): Deep Learning and Convolutional
Neural Networks for Medical Image Computing. ACVPR. Springer, Cham (2017).
https://doi.org/10.1007/978-3-319-42999-1
16. Ozcan, A., Bilenca, A., Desjardins, A.E., Bouma, B.E., Tearney, G.J.: Speckle
reduction in optical coherence tomography images using digital filtering. JOSA A
24(7), 1901–1910 (2007)
17. Ronneberger, O., Fischer, P., Brox, T.: U-Net: convolutional networks for biomed-
ical image segmentation. In: Navab, N., Hornegger, J., Wells, W.M., Frangi, A.F.
(eds.) MICCAI 2015. LNCS, vol. 9351, pp. 234–241. Springer, Cham (2015).
https://doi.org/10.1007/978-3-319-24574-4 28
18. Simonyan, K., Zisserman, A.: Very deep convolutional networks for large-scale
image recognition. arXiv preprint arXiv:1409.1556 (2014)
19. Suganya, P., Gayathri, S., Mohanapriya, N.: Survey on image enhancement tech-
niques. Int. J. Comput. Appl. Technol. Res. 2(5), 623–627 (2013)
20. Szegedy, C., Ioffe, S., Vanhoucke, V., Alemi, A.A.: Inception-v4, inception-resnet
and the impact of residual connections on learning. In: Thirty-First AAAI Confer-
ence on Artificial Intelligence (2017)
21. van Velthoven, M.E., Faber, D.J., Verbraak, F.D., van Leeuwen, T.G., de Smet,
M.D.: Recent developments in optical coherence tomography for imaging the retina.
Prog. Retin. Eye Res. 26(1), 57–77 (2007)
22. Venhuizen, F.G., et al.: Robust total retina thickness segmentation in optical
coherence tomography images using convolutional neural networks. Biomed. Opt.
Express 8(7), 3292–3316 (2017)
23. Wang, X., et al.: ESRGAN: enhanced super-resolution generative adversarial net-
works. In: Proceedings of the European Conference on Computer Vision (ECCV)
(2018)
24. Wang, Z., Chen, J., Hoi, S.C.: Deep learning for image super-resolution: a survey.
arXiv preprint arXiv:1902.06068 (2019)
25. Wang, Z., Bovik, A.C., Sheikh, H.R., Simoncelli, E.P., et al.: Image quality assess-
ment: from error visibility to structural similarity. IEEE Trans. Image Process.
13(4), 600–612 (2004)
26. Yang, W., Zhang, X., Tian, Y., Wang, W., Xue, J.H., Liao, Q.: Deep learning
for single image super-resolution: a brief review. IEEE Trans. Multimed. 21(12),
3106–3121 (2019)