RAISR：谷歌提出的速度与质量兼备的超分辨率技术

需积分: 50 119 浏览量更新于2024-07-19 2 收藏 8.88MB PDF 举报

“RAISR：快速且精确的图像超分辨率技术，由Yaniv Romano, John Isidoro和Peyman Milanfar（IEEE Fellow）提出。该技术旨在以极快的速度提升图像的质量和分辨率，同时在性能上比现有的如A+等算法提高10到100倍。” 在计算机视觉和图像处理领域，超分辨率（Super-Resolution）是一项关键的技术，其目的是通过增加像素数量和改善图像质量将低分辨率图像转换为高分辨率图像。单图像超分辨率（Single Image Super-Resolution, SISR）问题则特别关注仅使用单个输入图像来生成高分辨率版本。传统的方法往往需要大量的训练数据——即低分辨率和对应的高分辨率图像对——来学习一组滤波器或映射函数，用于将新的低分辨率图像转化为高分辨率。 RAISR（Rapid and Accurate Image Super-Resolution）方法与众不同之处在于它的运行速度。与现有的最佳竞争方法相比，RAISR的运行时间快了两个数量级以上，这意味着它能在保持甚至超越现有最先进技术水平的同时，实现显著的性能提升。这种高效性使得实时或大规模的图像超分辨率应用成为可能。此外，RAISR还涉及图像锐化和对比度增强。通过对模糊图像中的细节进行放大来提高视觉质量，它能以极其高效的方式生成比原始模糊图像更清晰的图像，而无需牺牲处理速度。这种方法扩展了RAISR的应用范围，不仅限于分辨率提升，还能用于改善图像的整体视觉效果。 RAISR通过创新的滤波器学习策略和优化的运行时间，实现了超分辨率领域的重大突破，它在提升图像质量和处理速度之间找到了一个理想的平衡点。对于需要高效处理大量图像的场景，例如视频流处理、监控系统或者移动设备上的图像应用，RAISR都是一种极具潜力的技术。同时，其在图像锐化方面的贡献也进一步增强了其在图像增强领域的实用性。

Fig. 3. Spatially varying learning scheme of four global ﬁlters, taking into consideration the internal structure of the bilinear.

(a) P1-Filter (b) P2-Filter (c) P3-Filter (d) P4-Filter

(e) P1-Magnitude spectrum (f) P2-Magnitude spectrum (g) P3-Magnitude spectrum (h) P4-Magnitude spectrum

Fig. 4. Visualization of the four global ﬁlters, corresponding to P1-P4 type of pixels, in the pixel domain (a-d), along with their magnitude

in the frequency domain (e-f), where the warmer the color, the larger the value. The ﬁlters are learned on Fig. 2 image.

on the LR image according to the type of the pixel, denoted by P1-P4. Since a convolution of two linear ﬁlters can

be uniﬁed into one ﬁlter (in our case, the ﬁrst is the bilinear and the second is the pre-learned one)

, we should

learn four different ﬁlters, corresponding to the four possible types of pixels, as demonstrated in Fig. 3.

The importance of this observation is illustrated in Fig. 4, which plots examples of actual learned ﬁlters, along

with their magnitude in the frequency domain. The obtained ﬁlters act like bandpass ﬁlters, amplifying the mid-

frequencies, and suppressing the high-frequencies (which contain aliasing components) of the interpolated image.

The learned ﬁlters have similar magnitude response (Fig. 4e-4h), but different phase response (Fig. 4a-4d), standing

This observation is a promising way to further speed up the algorithm and reduce the overall complexity.

Fig. 5. Applying the four spatially varying pre-learned ﬁlters on a LR image.

in agreement with the four different shifted versions of the interpolation kernels.

On the application side, similarly to the core/naive upscaling idea, we ﬁrst upscale the LR image using the bilinear

interpolator. Then, differently from the naive approach, we apply the pre-learned ﬁlters according to the type of

the pixel, followed by an aggregation step that simply combines the outcome of the ﬁltered patches (resulting in a

pixel) to an image. This process is illustrated in Fig. 5.

Notice that a similar observation holds for upscaling by any other integer factor s. For example, upscaling by

a factor of 3 implies that we should learn 9 ﬁlters, one per each pixel-type. Similarly, when upscaling by a factor

of 4, there are 16 types of pixels. As already mentioned, in order to keep the ﬂow of the explanations, we will

concentrate on the 2× scenario since the generalization to other scaling factors is straightforward.

IV. RAISR: HASHING-BASED LEARNING AND UPSCALING

Generally speaking, the global image ﬁltering is fast and cheap, as it implies the application of one ﬁlter per

patch. Since the learning scheme reduces the Euclidean distance between the HR and the interpolated version

of the LR images, the global ﬁltering has the ability to improve the restoration performance of various linear

upscaling methods. However, the global approach described so far is weaker than the state-of-the-art algorithms,

e.g., sparsity-based methods [8]–[11] or the neural networks based ones [16] that build upon large amount of

parameters, minimizing highly nonlinear cost functions. In contrast to these methods, the global approach is not

adaptive to the content of the image, and its learning stage estimates only a small amount of parameters.

Adaptivity to the image content can be achieved by dividing the image patches into clusters, and constructing

an appropriate ﬁlter per each cluster (e.g. as done in [10], [11]). However, the clustering implies the increase

of the overall complexity of the algorithm, which is an outgrowth that we want to avoid. Therefore, instead of

applying ”expensive” clustering (e.g. K-means [32], GMM [33], [34], dictionary learning [8], [9], [11], [14]–[18]),

we suggest using an efﬁcient hashing approach, leading to adaptive ﬁltering that keeps the low complexity of the

linear ﬁltering. More speciﬁcally, the local adaptivity is achieved by dividing the image patches into groups (called

”buckets”) based on an informative and ”cheap” geometry measures, which utilize the statistics of the gradients (a

剩余30页未读，继续阅读

kola

粉丝: 0
资源: 3

RAISR：谷歌提出的速度与质量兼备的超分辨率技术

超分辨率重建SRCNN(GUI界面)

超分辨率图像重建常用训练集1

Python-通过深度空间特征变换恢复图像超分辨率的真实纹理CVPR2018

相信大家一定都知道大名鼎鼎的 ascii 码,这次给你的任务是输入数字(表示 ascii 码

vnc libray

recover database '/home/dmdbms/data/DAMENG/dm.ini' from backupset '/home/dmdbms/data/DAMENG/bak/DB_DAMENG_INCREMENT_2023_04_26_22_30_56'; recover database '/home/dmdbms/data/DAMENG/dm.ini' from backupset '/home/dmdbms/data/DAMENG/bak/DB_DAMENG_INCREMENT_2023_04_26_22_30_56';

i2 notebook

restore database '/home/dmdbms/data/DAMENG/dm.ini' from backupset '/home/dmdbms/data/DAMENG/bak/DB_DAMENG_INCREMENT_2023_04_26_22_30_56' with backupdir '/home/dmdbms/data/DAMENG/bak/DB_DAMENG_FULL_2023_04_15_01_00_54';

c# secsgem报文解析为sml

最新资源