改进的显著性层级模型在物体识别中的应用
需积分: 10 148 浏览量
更新于2024-08-26
收藏 2.4MB PDF 举报
"物体识别的梯度层次模型是2012年国际小波分析与模式识别会议上的一篇研究论文，作者包括WEI-BIN YANG、BINFANG、ZHAO-WEI SHANG和BOLIN，他们都来自中国重庆大学计算机科学学院。这篇论文探讨了如何利用图像显著性来提升物体识别的效率和准确性。"
在物体识别领域，梯度层次模型是一种重要的技术，它试图模仿人类视觉选择性注意力机制，找出输入图像中最显著的部分。图像显著性（imagesaliency）计算是这一过程的关键，它旨在突出那些在背景中相对突出的特征，以便于后续的识别步骤。传统的显著性模型被修改以增强其鲁棒性，使其在复杂场景下也能准确估计图像的显著区域。
论文提出了一种新的显著性层次模型，将视觉显著性检测方法与层次最大化模型相结合。这种结合方式可以提供更丰富的视觉信息，对分类过程非常有帮助。层次化的方法允许模型逐步细化地处理图像信息，从全局到局部，逐层聚焦于潜在的物体特征。
实验结果证明，改进的显著性模型能够提取出更精确的显著区域，而提出的显著性层次模型在性能上超越了传统的层次最大化模型。这一发现强调了在物体识别过程中考虑图像显著性的重要性，并且可能对深度学习和计算机视觉领域的物体检测算法有所启发。
关键词包括：图像显著性、视觉皮层（模拟人类视觉系统）、层次模型。这些关键词表明，该研究不仅关注技术应用，还涉及到对人类视觉理解的借鉴，以及通过多层次的处理来优化识别效果。这为未来的物体识别研究提供了新的思路和可能的优化方向。
Proceedings
of
the 2012 International Conference on Wavelet Analysis and Pattern Recognition, Xian, 15-17 July, 2012
A SALIENT
HIERARCHICAL
MODEL
FOR
OBJECT
RECOGNITION
WEI-BIN
YANG,
BIN
FANG,
ZHAO-WEI
SHANG, BO
LIN
School
of
Computer Science, Chongqing University, Chongqing, China
E-MAIL:
ywb@cqu.edu.cn
Abstract:
Image saliency attempts to describe the most conspicuous
part
in an input image by mimicking human visual selective
attention mechanism. Naturally, it could be adopted for
improving object recognition. To demonstrate the
effectiveness of saliency in object recognition, this
paper
proposes a salient hierarchical modeL First, the traditional
saliency model is modified for more robust saliency estimation.
Second, the visual saliency detection method is combined with
the Hierarchical Maximization model to provide more useful
visual information for classification. Experimental results
show
that
the improved saliency model extracts more accurate
conspicuity, and the proposed salient hierarchical model
outperforms Hierarchical Maximization modeL
Keywords:
Image saliency; visual cortex; hierarchical model; object
recognition
1. Introduction
To learn how humans look and recognize is an
important issue in computer vision and pattern recognition.
Two main involving research topics are visual saliency
detection and object recognition. Mostly, we develop
related research work in different ways. However, since
both two tasks are inspired by human visual system and
visual cortex, it is reasonable to believe that the research
achievement may benefit each other. Therefore, a robust
saliency model and an effective combination may be the
key for attention based object recognition.
Visual saliency is believed to drive human fixation
behavior during free viewing by attracting visual attention
in a bottom-up way. Moreover, saliency also appears to
determine which details humans find interesting in visual
scenes [1]. The most influential computational framework
for estimating visual saliency is proposed by Itti et al. [2],
which implemented and further developed the
physiologically inspired saliency-based model
of
visual
attention introduced by Koch and Ullman [3]. Itti's saliency
model first computes feature maps for color, intensity and
orientation using a center-surround operator across different
978·1-4673·1535·7/121$31.00 ©2012 Crown
scales, and then generates the saliency map by
normalization and summation on these feature maps.
Achanta et al. [4] used features
of
color and luminance to
detect salient region with well-defmed boundaries.
Goferman et al. [5] presented a saliency detector by
computing the dissimilarity between different image
patches over four scales. Cheng et al. [6] proposed a global
method to detect visual saliency by measuring the
dissimilarity between different image regions, which
obtained excellent performance on salient object detection.
Hou et al. [7] proposed an image descriptor, denoted image
signature, to approximate the foreground
of
an image using
the Discrete Cosine Transform and Inverse Discrete Cosine
Transform.
In addition, based on our knowledge
of
visual vortex,
many studies focus on biologically plausible method for
object class recognition. Recent work by Serre et al. [8]
proposed a computational model (Hierarchical
Maximization,
HMAX)
based on the feedforward path
of
object recognition in cortex that accounts for the first
100-200 milliseconds
of
processing in the ventral stream
of
primate visual cortex [9]. HMAX model obtains promising
results on some
of
the standard classification datasets.
Mutch et al. [10] improved HMAX model by incorporating
some additional biologically-motivated properties, such as
sparsity and localized intermediated-level features.
To prove visual saliency is useful for object
recognition, Riesenhuber et al. [11] applied Itti's saliency
model with SIFT descriptor. Han et al. [12] combined
attention and recognition by replacing the first layer
of
the
HMAX architecture with a saliency network. In this paper,
we attempt to provide a new view in another way. We use
saliency model to guide the learning process and to form
the principle
of
choosing training samples in HMAX
model.
The rest
of
this paper is organized as follows. Section
2 introduces the proposed salient hierarchical model in
detail. Section 3 evaluates the performance
of
the improved
saliency model and the proposed salient hierarchical model.
Conclusions are given in Section 4.
244
下载后可阅读完整内容，剩余5页未读，立即下载
2021-09-25 上传
2024-09-21 上传
点击了解资源详情
2024-05-18 上传
点击了解资源详情
点击了解资源详情
104 浏览量
点击了解资源详情
点击了解资源详情
weixin_38660624
- 粉丝: 3
我的内容管理
展开
最新资源
- 前端技术分享：全面的JavaScript 示例教程
- Ruby项目active_admin_sample部署与运行指南
- 重播扑克Replay Bankroll Chart-crx插件使用指南
- Android基础实例解析：天气、地图、音乐播放器等源码
- JCms v1.5.3：Asp.NET内容管理系统助力电子政务与校园门户建设
- Apache Beam MySQL连接器：轻松读取MySQL数据库数据
- 深入解析词云技术在网络文本分析中的应用
- Node.js环境下hyperdb分布式数据库的应用与扩展
- 网络性能测试与评估：tp-at-arq_redes_infnet深入分析
- 掌握Python数据结构：问题集练习指南
- 基于BART模型的神经故事生成技术研究
- 前端美化神器：Ion.RangeSlider实现及示例解析
- C++实现3DES与Base64加解密方法示例
- 探索Dodger.js：Vimscript下的JavaScript开发利器
- Python打包服务器项目实现自动化发布与一键部署
- Python实践教程：HuohuaTest01压缩包子文件解析
