GithubHelp home page GithubHelp logo

raisr's Introduction

实现论文:Y. Romano, J. Isidoro, and P. Milanfar, "RAISR: Rapid and Accurate Image Super Resolution"

环境:Matlab R2016a

直接运行 runRAISRTest.m 就能完成训练和测试的全过程。可以直接更改 scale 和 LR_patch_size 的值,训练代码能根据给定的这两个参数来进行新的训练。

RAISR 的训练过程较长,因为要从插值后的 HR 图像中提取 patch,大概需要 3-4 小时。训练好的结果已经保存在 RAISR_result 文件夹下,直接把 runRAISRTest.m 的 22、23 行注释掉就能直接进行测试。

结果

用的 patch_size 为 11,scale 为 3,从训练图像中提取出了 7000w 的 patch 进行训练。

计算时间在 MacOS, 2 GHz Intel Core i5, 8 GB 1867 MHz LPDDR3 下进行测试得出;其中用的 Bicubic 方法为自己实现的,没有实现 Antialiasing 功能

Set14 Bicubic RAISR
PSNR SSIM PSNR SSIM Time
baboon 21.53 0.54 23.39 0.60 28.01s
barbara 24.74 0.76 26.48 0.78 49.26s
bridge 23.42 0.66 24.85 0.70 31.38s
coastguard 25.20 0.63 26.97 0.66 12.16s
comic 22.55 0.72 24.09 0.76 10.05s
face 31.66 0.78 33.61 0.82 8.62s
flowers 26.88 0.81 28.43 0.84 21.98s
foreman 28.69 0.90 31.24 0.92 12.25s
lenna 31.40 0.86 33.15 0.88 29.24s
man 26.28 0.75 27.88 0.79 32.24s
monarch 29.17 0.92 30.96 0.93 50.26s
pepper 29.47 0.86 31.65 0.88 31.51s
ppt3 23.29 0.89 24.82 0.91 42.16s
zebra 26.69 0.81 28.65 0.84 27.63s
AVERAGE 26.50 0.78 28.30 0.81 28.13s

结果也与论文给出的测试结果相近:

a4

选部分结果进行展示:

r7

baboon 的鼻子处的噪点更平滑了。

r6

flowers 的所有过渡的边缘部分都变得没有那么锐利,过渡更自然。

r5

foreman 的结果有了很大的提高,主要是因为后方墙壁的边缘很好的平滑掉了,SF 的结果中墙壁的边缘有很多噪点,而 RAISR 的中不会出现。

r4

r3

可见 RAISR 的结果比 SF 看起来都更加自然和平滑,并且基本都有更高的 PSNR 值。

raisr's People

Watchers

James Cloos avatar

Recommend Projects

  • React photo React

    A declarative, efficient, and flexible JavaScript library for building user interfaces.

  • Vue.js photo Vue.js

    🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.

  • Typescript photo Typescript

    TypeScript is a superset of JavaScript that compiles to clean JavaScript output.

  • TensorFlow photo TensorFlow

    An Open Source Machine Learning Framework for Everyone

  • Django photo Django

    The Web framework for perfectionists with deadlines.

  • D3 photo D3

    Bring data to life with SVG, Canvas and HTML. 📊📈🎉

Recommend Topics

  • javascript

    JavaScript (JS) is a lightweight interpreted programming language with first-class functions.

  • web

    Some thing interesting about web. New door for the world.

  • server

    A server is a program made to process requests and deliver data to clients.

  • Machine learning

    Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.

  • Game

    Some thing interesting about game, make everyone happy.

Recommend Org

  • Facebook photo Facebook

    We are working to build community through open source technology. NB: members must have two-factor auth.

  • Microsoft photo Microsoft

    Open source projects and samples from Microsoft.

  • Google photo Google

    Google ❤️ Open Source for everyone.

  • D3 photo D3

    Data-Driven Documents codes.