随机森林：高效图像分类与分割算法

4星 · 超过85%的资源需积分: 50 143 浏览量更新于2024-09-24 8 收藏 2.59MB PDF 举报

随机森林图像分类分割是一种高效且强大的计算机视觉技术，由Jamie Shotton、Matthew Johnson和Roberto Cipolla在他们的研究中提出。该方法的主要创新在于所谓的"SemanticTextonForests"，这是一种基于决策树的低级特征表示，直接作用于像素级别，无需进行昂贵的滤波器银行响应或局部特征描述计算。这使得模型在训练和测试阶段都具有极高的速度优势，尤其相对于传统的k-means聚类和基于特征描述符的最近邻分配。在SemanticTextonForests中，每个决策树节点提供了两个关键信息：(i)隐含的语义纹理簇层次结构，这些簇是根据像素值和周围环境自动组织的；(ii)对局部区域的明确分类估计。这种直接操作像素的方式减少了对高级特征工程的依赖，提高了算法的效率。此外，研究者还提出了"Bag of Semantic Textons"的概念，这是一种结合了图像区域内的语义纹理直方图和区域先验类别分布的方法。对于图像分类，它将整个图像作为分析单位，而为了图像分割，它则应用到局部矩形区域上。这种方法同时考虑了局部特征的统计分布（直方图）和全局上下文信息（先验类别分布），从而增强了分割的准确性。随机森林图像分类分割的优势在于其速度、精度和灵活性，尤其是在处理大规模图像数据集时，其性能表现优秀。通过利用这些低维、直观的特征和有效的处理策略，它在图像理解和处理任务中展现出了强大的实用价值。与传统方法相比，它简化了预处理步骤，并能够在实时或近实时的场景中提供高效的结果。这项工作为图像分析领域的研究者和工程师提供了一种新的工具，促进了计算机视觉技术的发展。

Semantic Texton Forests for Image Categorization and Segmentation

Jamie Shotton

†

Matthew Johnson

Roberto Cipolla

†

Toshiba Corporate R&D Center

Kawasaki, Japan

Department of Engineering

University of Cambridge, UK

Abstract

We propose semantic texton forests, efﬁcient and pow-

erful new low-level features. These are ensembles of deci-

sion trees that act directly on image pixels, and therefore do

not need the expensive computation of ﬁlter-bank responses

or local descriptors. They are extremely fast to both train

and test, especially compared with k-means clustering and

nearest-neighbor assignment of feature descriptors. The

nodes in the trees provide (i) an implicit hierarchical clus-

tering into semantic textons, and (ii) an explicit local clas-

siﬁcation estimate. Our second contribution, the bag of se-

mantic textons, combines a histogram of semantic textons

over an image region with a region prior category distri-

bution. The bag of semantic textons is computed over the

whole image for categorization, and over local rectangu-

lar regions for segmentation. Including both histogram and

region prior allows our segmentation algorithm to exploit

both textural and semantic context. Our third contribution

is an image-level prior for segmentation that emphasizes

those categories that the automatic categorization believes

to be present. We evaluate on two datasets including the

very challenging VOC 2007 segmentation dataset. Our re-

sults signiﬁcantly advance the state-of-the-art in segmenta-

tion accuracy, and furthermore, our use of efﬁcient decision

forests gives at least a ﬁve-fold increase in execution speed.

1. Introduction

This paper introduces semantic texton forests, and

demonstrates their use for image categorization and seman-

tic segmentation; see Figure

1. Our aim is to show that one

can build powerful texton codebooks without computing ex-

pensive ﬁlter-banks or descriptors, and without performing

costly k-means clustering and nearest-neighbor assignment.

Semantic texton forests (STFs) fulﬁll both criteria. They

are randomized decision forests that use only simple pixel

comparisons on local image patches, performing both an

implicit hierarchical clustering into semantic textons and

an explicit local classiﬁcation of the patch category. Our

results show that STFs improve the state-of-the-art in both

Figure 1. Semantic texton forests. (a) Test image, with ground

truth in-set. Semantic texton forests very efﬁciently compute (b) a

set of semantic textons per pixel and (c) a rough local segmenta-

tion prior. Our algorithm uses both textons and priors as features to

give coherent semantic segmentation (d), and even ﬁnds the build-

ing unmarked in the ground truth. Colors show texton indices in

(b), but categories corresponding to the ground truth in (c) and (d).

quantitative performance and execution speed.

We look at two applications of STFs: image categoriza-

tion (inferring the object categories present in an image)

and semantic segmentation (dividing the image into coher-

ent regions and simultaneously categorizing each region).

To these ends, we propose the bag of semantic textons. This

is computed over a given image region, and extends the bag

of words model [ 4] by combining a histogram of the hierar-

chical semantic textons with a region prior category distri-

bution. By considering the image as a whole, we obtain a

highly discriminative descriptor for categorization. For seg-

mentation, we use many local rectangular regions and build

a second randomized decision forest that achieves efﬁcient

and accurate segmentation.

Inferring the correct segmentation depends on local im-

age information that can often be ambiguous. The global

statistics of the image, however, are more discriminative

and may be sufﬁcient to accurately estimate the image cate-

gorization. We therefore investigate how categorization can

下载后可阅读完整内容，剩余7页未读，立即下载

newplan1986

粉丝: 0
资源: 1

随机森林：高效图像分类与分割算法

三种方法实现图像的分割（含程序源码）

OpenCV使用随机森林实现对图像分类

多光谱图像多变量图像分析 GUI：多光谱图像数据集（PCA、Simplisma、MCR、分类）MIA 的 GUI。-matlab开发

利用随机森林-光学图像分割

随机森林图像matlab代码-SRFAff_v1:用于价格分割的结构化随机森林

随机森林图像matlab代码-brainseg:训练完全混合神经网络进行脑分割

随机森林图像matlab代码-Hierarchy-Tree-segmentation-:层次树分割

随机森林图像matlab代码-MS-challenge-2016:多发性硬化病变分割

随机森林图像matlab代码-StepForest:使用局部强度和纹理特征分割结肠组织学图像中的腺体的机器学习方法

image-segmentation-classification:此仓库包含一个Jupyter Notebook，该笔记本测试使用随机森林分类器对高分辨率NAIP图像进行分割和分类的方法

最新资源