基于TI-DMC642的显著性物体检测  

Salient object detection based on TI-DMC642 DSP platform

作　　者：李璇[1] 宁雅青[2] 陈劲松[1] 周博磊[2] 童善保[3] 

机构地区：[1]上海交通大学生命科学技术学院,上海200240 [2]上海交通大学电子信息与电气工程学院,上海200240 [3]上海交通大学Med-X研究院,上海200030

出　　处：《计算机应用》2010年第A01期298-301,共4页journal of Computer Applications

基　　金：德州仪器创新计划项目

摘　　要：基于硬件平台的显著性物体检测技术应用到生产生活中,便可以通过便携设备,实现诸如盲人的视觉导航、交通系统中的路牌识别等问题,具有很高的实用价值。基于TI-DMC642平台,采用谱残差的方法,解决图像分辨率调整、IFFT编码、滤波和阈值化等问题后,可以最终得到效果不错的显著性物体检测并实现在PC机上很难做到的动态实时显示。It is of great importance to extract salient object from a background regardless of the details of the object. Although the detection of salient objects based on personal computer has made rapid progress in robot vision and pattern recognition, there is not a system based on DSP platform for salience detection. On the other hand, a DSP-based portable system can be very useful in the vision guidance for the blind, the recognition of street names, and etc. By use of the model of spectral residual, after solving technical details, such as the modulation of the spatial resolution, the coding of IFFT, filtering and thresholding, the TI-DMC 642-based salient object detection has a powerful performance and processes the images in real time.

关 键 词：谱残差 滤波 显著性图 硬件实现 实时显示 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]