<div id="readability-page-1" class="page"><div id="content">
						<article id="post-6060">
				<div>
OpenCV Python에서 화질 손실을 최소화하여 이미지를 확대하기 위해 Super Resolution을 사용하는 방법을 소개하고 있습니다.



2022. 11. 5 최초작성



2022. 11. 6 최종작성







OpenCV에선 미리 학습된 Super Resolution 딥러닝 모델을 사용하여 이미지를 확대할 수 있는 방법을 제공합니다.



본 포스팅에서는 실제 사용하는 방법을 소개합니다.







1. 우선 contrib 모듈이 포함된 OpenCV를 설치해야 합니다. 기존에 opencv-python 패키지를 설치한적이 있다면 삭제하고 설치하세요. pip list 명령으로 설치된 패키지 목록을 확인할 수 있습니다.



<pre><code>pip install opencv-contrib-python</code></pre>







2. 미리 학습되어 배포되고 있는 모델을 다운로드합니다.



EDSR 가장 좋은 결과를 얻을 수 있습니다. 모델이 제일 크고 추론 속도가 가장 느립니다.



<a href="https://github.com/Saafke/EDSR_Tensorflow/tree/master/models">https://github.com/Saafke/EDSR_Tensorflow/tree/master/models</a>



ESPCN, FSRCNN 작은 모델로 추론 속도가 빠르면서 적당히 좋은 결과를 보여줍니다. 실시간 비디오에 적용할 수 있습니다.



ESPCN <a href="https://github.com/fannymonori/TF-ESPCN/tree/master/export">https://github.com/fannymonori/TF-ESPCN/tree/master/export</a>



FSRCNN <a href="https://github.com/Saafke/FSRCNN_Tensorflow/tree/master/models">https://github.com/Saafke/FSRCNN_Tensorflow/tree/master/models</a>



LapSRN 중간 크기의 모델입니다.



<a href="https://github.com/fannymonori/TF-LapSRN/tree/master/export">https://github.com/fannymonori/TF-LapSRN/tree/master/export</a>







몇 배 확대할지에 따라 다운받을 모델을 선택해야 합니다.



예를 들어 ESPCN을 대상으로 포스트를 진행하도록 하겠습니다. 모델에는 N배 확대를 위한 모델 이름에 xN이 적혀있습니다. 따라서 2배 확대를 위해선 ESPCN_x2.pb를 선택하면 됩니다. ESPCN_x2.pb를 클릭합니다.



<figure><img data-lazy-fallback="1" data-attachment-id="6061" data-permalink="https://webnautes.blog/6060/word-image-6060-1/" data-orig-file="https://i0.wp.com/webnautes.blog/wp-content/uploads/2023/09/word-image-6060-1.png?fit=1550%2C804&amp;ssl=1" data-orig-size="1550,804" data-comments-opened="1" data-image-meta="{&quot;aperture&quot;:&quot;0&quot;,&quot;credit&quot;:&quot;&quot;,&quot;camera&quot;:&quot;&quot;,&quot;caption&quot;:&quot;&quot;,&quot;created_timestamp&quot;:&quot;0&quot;,&quot;copyright&quot;:&quot;&quot;,&quot;focal_length&quot;:&quot;0&quot;,&quot;iso&quot;:&quot;0&quot;,&quot;shutter_speed&quot;:&quot;0&quot;,&quot;title&quot;:&quot;&quot;,&quot;orientation&quot;:&quot;0&quot;}" data-image-title="word-image-6060-1" data-image-description="" data-image-caption="" data-medium-file="https://i0.wp.com/webnautes.blog/wp-content/uploads/2023/09/word-image-6060-1.png?fit=300%2C156&amp;ssl=1" data-large-file="https://i0.wp.com/webnautes.blog/wp-content/uploads/2023/09/word-image-6060-1.png?fit=1024%2C531&amp;ssl=1" decoding="async" fetchpriority="high" width="1200" height="622" src="https://i0.wp.com/webnautes.blog/wp-content/uploads/2023/09/word-image-6060-1.png?resize=1200%2C622&amp;ssl=1" alt="" sizes="(max-width: 1200px) 100vw, 1200px" data-recalc-dims="1" data-old-src="https://i0.wp.com/webnautes.blog/wp-content/uploads/2023/09/word-image-6060-1.png?resize=1200%2C622&amp;is-pending-load=1#038;ssl=1" data-lazy-srcset="https://i0.wp.com/webnautes.blog/wp-content/uploads/2023/09/word-image-6060-1.png?w=1550&amp;ssl=1 1550w, https://i0.wp.com/webnautes.blog/wp-content/uploads/2023/09/word-image-6060-1.png?resize=300%2C156&amp;ssl=1 300w, https://i0.wp.com/webnautes.blog/wp-content/uploads/2023/09/word-image-6060-1.png?resize=1024%2C531&amp;ssl=1 1024w, https://i0.wp.com/webnautes.blog/wp-content/uploads/2023/09/word-image-6060-1.png?resize=768%2C398&amp;ssl=1 768w, https://i0.wp.com/webnautes.blog/wp-content/uploads/2023/09/word-image-6060-1.png?resize=1536%2C797&amp;ssl=1 1536w, https://i0.wp.com/webnautes.blog/wp-content/uploads/2023/09/word-image-6060-1.png?resize=1200%2C622&amp;ssl=1 1200w"></figure>







3. Download를 클릭하면 모델이 다운로드됩니다. 파이썬 코드를 작성할 폴더로 복사해두세요.



<figure><img data-lazy-fallback="1" data-attachment-id="6062" data-permalink="https://webnautes.blog/6060/word-image-6060-2/" data-orig-file="https://i0.wp.com/webnautes.blog/wp-content/uploads/2023/09/word-image-6060-2.png?fit=1558%2C910&amp;ssl=1" data-orig-size="1558,910" data-comments-opened="1" data-image-meta="{&quot;aperture&quot;:&quot;0&quot;,&quot;credit&quot;:&quot;&quot;,&quot;camera&quot;:&quot;&quot;,&quot;caption&quot;:&quot;&quot;,&quot;created_timestamp&quot;:&quot;0&quot;,&quot;copyright&quot;:&quot;&quot;,&quot;focal_length&quot;:&quot;0&quot;,&quot;iso&quot;:&quot;0&quot;,&quot;shutter_speed&quot;:&quot;0&quot;,&quot;title&quot;:&quot;&quot;,&quot;orientation&quot;:&quot;0&quot;}" data-image-title="word-image-6060-2" data-image-description="" data-image-caption="" data-medium-file="https://i0.wp.com/webnautes.blog/wp-content/uploads/2023/09/word-image-6060-2.png?fit=300%2C175&amp;ssl=1" data-large-file="https://i0.wp.com/webnautes.blog/wp-content/uploads/2023/09/word-image-6060-2.png?fit=1024%2C598&amp;ssl=1" decoding="async" width="1200" height="701" src="https://i0.wp.com/webnautes.blog/wp-content/uploads/2023/09/word-image-6060-2.png?resize=1200%2C701&amp;ssl=1" alt="" sizes="(max-width: 1200px) 100vw, 1200px" data-recalc-dims="1" data-old-src="https://i0.wp.com/webnautes.blog/wp-content/uploads/2023/09/word-image-6060-2.png?resize=1200%2C701&amp;is-pending-load=1#038;ssl=1" data-lazy-srcset="https://i0.wp.com/webnautes.blog/wp-content/uploads/2023/09/word-image-6060-2.png?w=1558&amp;ssl=1 1558w, https://i0.wp.com/webnautes.blog/wp-content/uploads/2023/09/word-image-6060-2.png?resize=300%2C175&amp;ssl=1 300w, https://i0.wp.com/webnautes.blog/wp-content/uploads/2023/09/word-image-6060-2.png?resize=1024%2C598&amp;ssl=1 1024w, https://i0.wp.com/webnautes.blog/wp-content/uploads/2023/09/word-image-6060-2.png?resize=768%2C449&amp;ssl=1 768w, https://i0.wp.com/webnautes.blog/wp-content/uploads/2023/09/word-image-6060-2.png?resize=1536%2C897&amp;ssl=1 1536w, https://i0.wp.com/webnautes.blog/wp-content/uploads/2023/09/word-image-6060-2.png?resize=1200%2C701&amp;ssl=1 1200w"></figure>







4. 다음 코드를 작성하여 실행합니다.



<pre><code>import cv2


# 이미지를 불러옵니다.
img = cv2.imread('cat.jpeg')


# Super Resolution을 하기 위해 2배 확대를 위한 ESPCN 모델을 사용합니다.
sr = cv2.dnn_superres.DnnSuperResImpl_create()
# 모델 파일의 이름을 적어주면 됩니다. 
sr.readModel('ESPCN_x2.pb')
# 모델 파일에 적힌 숫자와 일치하도록 적어줘야 합니다.
sr.setModel("espcn",2)
# img를 입력 받아 결과로 result1를 돌려줍니다. 
result1 = sr.upsample(img)


# OpenCV에서 제공하는 보간법을 사용해봅니다.
result2 = cv2.resize(img, dsize=None, fx=2, fy=2)

# 입력 이미지와 Super Resolution, OpenCV 보간법 적용 결과를 화면에 보여줍니다.
cv2.imshow('origianl', img)
cv2.imshow('result - ESPCN', result1)
cv2.imshow('result - Opencv', result2)
cv2.waitKey(0)</code></pre>







5. 실행 결과입니다. 어느쪽이 더 나은 결과인지 판단하기 쉽지 않네요.



<figure><img data-lazy-fallback="1" data-attachment-id="6063" data-permalink="https://webnautes.blog/6060/word-image-6060-3/" data-orig-file="https://i0.wp.com/webnautes.blog/wp-content/uploads/2023/09/word-image-6060-3.png?fit=460%2C524&amp;ssl=1" data-orig-size="460,524" data-comments-opened="1" data-image-meta="{&quot;aperture&quot;:&quot;0&quot;,&quot;credit&quot;:&quot;&quot;,&quot;camera&quot;:&quot;&quot;,&quot;caption&quot;:&quot;&quot;,&quot;created_timestamp&quot;:&quot;0&quot;,&quot;copyright&quot;:&quot;&quot;,&quot;focal_length&quot;:&quot;0&quot;,&quot;iso&quot;:&quot;0&quot;,&quot;shutter_speed&quot;:&quot;0&quot;,&quot;title&quot;:&quot;&quot;,&quot;orientation&quot;:&quot;0&quot;}" data-image-title="word-image-6060-3" data-image-description="" data-image-caption="" data-medium-file="https://i0.wp.com/webnautes.blog/wp-content/uploads/2023/09/word-image-6060-3.png?fit=263%2C300&amp;ssl=1" data-large-file="https://i0.wp.com/webnautes.blog/wp-content/uploads/2023/09/word-image-6060-3.png?fit=460%2C524&amp;ssl=1" decoding="async" width="460" height="524" src="https://i0.wp.com/webnautes.blog/wp-content/uploads/2023/09/word-image-6060-3.png?resize=460%2C524&amp;ssl=1" alt="" sizes="(max-width: 460px) 100vw, 460px" data-recalc-dims="1" data-old-src="https://i0.wp.com/webnautes.blog/wp-content/uploads/2023/09/word-image-6060-3.png?resize=460%2C524&amp;is-pending-load=1#038;ssl=1" data-lazy-srcset="https://i0.wp.com/webnautes.blog/wp-content/uploads/2023/09/word-image-6060-3.png?w=460&amp;ssl=1 460w, https://i0.wp.com/webnautes.blog/wp-content/uploads/2023/09/word-image-6060-3.png?resize=263%2C300&amp;ssl=1 263w"></figure>



<figure><img data-lazy-fallback="1" data-attachment-id="6064" data-permalink="https://webnautes.blog/6060/word-image-6060-4/" data-orig-file="https://i0.wp.com/webnautes.blog/wp-content/uploads/2023/09/word-image-6060-4.png?fit=1816%2C978&amp;ssl=1" data-orig-size="1816,978" data-comments-opened="1" data-image-meta="{&quot;aperture&quot;:&quot;0&quot;,&quot;credit&quot;:&quot;&quot;,&quot;camera&quot;:&quot;&quot;,&quot;caption&quot;:&quot;&quot;,&quot;created_timestamp&quot;:&quot;0&quot;,&quot;copyright&quot;:&quot;&quot;,&quot;focal_length&quot;:&quot;0&quot;,&quot;iso&quot;:&quot;0&quot;,&quot;shutter_speed&quot;:&quot;0&quot;,&quot;title&quot;:&quot;&quot;,&quot;orientation&quot;:&quot;0&quot;}" data-image-title="word-image-6060-4" data-image-description="" data-image-caption="" data-medium-file="https://i0.wp.com/webnautes.blog/wp-content/uploads/2023/09/word-image-6060-4.png?fit=300%2C162&amp;ssl=1" data-large-file="https://i0.wp.com/webnautes.blog/wp-content/uploads/2023/09/word-image-6060-4.png?fit=1024%2C551&amp;ssl=1" decoding="async" loading="lazy" width="1200" height="646" src="https://i0.wp.com/webnautes.blog/wp-content/uploads/2023/09/word-image-6060-4.png?resize=1200%2C646&amp;ssl=1" alt="" sizes="(max-width: 1200px) 100vw, 1200px" data-recalc-dims="1" data-old-src="https://i0.wp.com/webnautes.blog/wp-content/uploads/2023/09/word-image-6060-4.png?resize=1200%2C646&amp;is-pending-load=1#038;ssl=1" data-lazy-srcset="https://i0.wp.com/webnautes.blog/wp-content/uploads/2023/09/word-image-6060-4.png?w=1816&amp;ssl=1 1816w, https://i0.wp.com/webnautes.blog/wp-content/uploads/2023/09/word-image-6060-4.png?resize=300%2C162&amp;ssl=1 300w, https://i0.wp.com/webnautes.blog/wp-content/uploads/2023/09/word-image-6060-4.png?resize=1024%2C551&amp;ssl=1 1024w, https://i0.wp.com/webnautes.blog/wp-content/uploads/2023/09/word-image-6060-4.png?resize=768%2C414&amp;ssl=1 768w, https://i0.wp.com/webnautes.blog/wp-content/uploads/2023/09/word-image-6060-4.png?resize=1536%2C827&amp;ssl=1 1536w, https://i0.wp.com/webnautes.blog/wp-content/uploads/2023/09/word-image-6060-4.png?resize=1200%2C646&amp;ssl=1 1200w"></figure>







정확한 비교를 해놓은 다음 참고 사이트에서 확인해보세요.



<figure></figure>







포스트에서 사용한 ESPCN의 경우만 링크에서 가져와 보면 다음과 같습니다. ESPCN을 사용한 경우가 OpenCV에서 사용하는 Bicubic 보간법보단 화질이 더 좋은 걸 볼 수 있습니다. ESPCN은 실시간으로 사용하기 위해 선택을 해본 것이라 화질이 덜 좋을 수 있습니다. 실시간이 필요없다면 더 좋은 화질을 얻을 수 있는 EDSR 모델을 고려해보세요.



<figure><img data-lazy-fallback="1" data-attachment-id="6065" data-permalink="https://webnautes.blog/6060/word-image-6060-5/" data-orig-file="https://i0.wp.com/webnautes.blog/wp-content/uploads/2023/09/word-image-6060-5.png?fit=1146%2C530&amp;ssl=1" data-orig-size="1146,530" data-comments-opened="1" data-image-meta="{&quot;aperture&quot;:&quot;0&quot;,&quot;credit&quot;:&quot;&quot;,&quot;camera&quot;:&quot;&quot;,&quot;caption&quot;:&quot;&quot;,&quot;created_timestamp&quot;:&quot;0&quot;,&quot;copyright&quot;:&quot;&quot;,&quot;focal_length&quot;:&quot;0&quot;,&quot;iso&quot;:&quot;0&quot;,&quot;shutter_speed&quot;:&quot;0&quot;,&quot;title&quot;:&quot;&quot;,&quot;orientation&quot;:&quot;0&quot;}" data-image-title="word-image-6060-5" data-image-description="" data-image-caption="" data-medium-file="https://i0.wp.com/webnautes.blog/wp-content/uploads/2023/09/word-image-6060-5.png?fit=300%2C139&amp;ssl=1" data-large-file="https://i0.wp.com/webnautes.blog/wp-content/uploads/2023/09/word-image-6060-5.png?fit=1024%2C474&amp;ssl=1" decoding="async" loading="lazy" width="1146" height="530" src="https://i0.wp.com/webnautes.blog/wp-content/uploads/2023/09/word-image-6060-5.png?resize=1146%2C530&amp;ssl=1" alt="" sizes="(max-width: 1146px) 100vw, 1146px" data-recalc-dims="1" data-old-src="https://i0.wp.com/webnautes.blog/wp-content/uploads/2023/09/word-image-6060-5.png?resize=1146%2C530&amp;is-pending-load=1#038;ssl=1" data-lazy-srcset="https://i0.wp.com/webnautes.blog/wp-content/uploads/2023/09/word-image-6060-5.png?w=1146&amp;ssl=1 1146w, https://i0.wp.com/webnautes.blog/wp-content/uploads/2023/09/word-image-6060-5.png?resize=300%2C139&amp;ssl=1 300w, https://i0.wp.com/webnautes.blog/wp-content/uploads/2023/09/word-image-6060-5.png?resize=1024%2C474&amp;ssl=1 1024w, https://i0.wp.com/webnautes.blog/wp-content/uploads/2023/09/word-image-6060-5.png?resize=768%2C355&amp;ssl=1 768w"></figure>







참고



<figure></figure>



<figure></figure>





</div>

			</article>
			
		</div></div>

멈춤보단 천천히라도

webnautes

OpenCV Python에서 화질 손실을 최소화하여 이미지를 확대하기 위해 Super Resolution을 사용하는 방법을 소개하고 있습니다. 2022. 11. 5 최초작성 2022. 11. 6 최종작성 OpenCV에선 미리 학습된 Super Resolution 딥러닝 모델을 사용하여 이미지를 확대할 수 있는 방법을 제공합니다. 본 포스팅에서는 실제 사용하는 방법을 소개합니다. 1. 우선 contrib 모듈이 포함된 OpenCV를 설치해야 합니다. 기존에 opencv-python 패키지를 설치한적이 있다면… 더 보기 »OpenCV Python – Super Resolution 예제
The post OpenCV Python – Super Resolution 예제 first appeared on 멈춤보단 천천히라도.

OpenCV Python – Super Resolution 예제

댓글