计算机科学与探索 ›› 2018, Vol. 12 ›› Issue (1): 163-170.DOI: 10.3778/j.issn.1673-9418.1609019

• 人工智能与模式识别 • 上一篇    

低对比度火焰图像增强和分割算法研究

韩铖惠,王慧琴+,胡  燕   

  1. 西安建筑科技大学 信息与控制工程学院,西安 710055
  • 出版日期:2018-01-01 发布日期:2018-01-09

Enhancement and Segmentation Algorithm Study for Low Contrast Fire Image

HAN Chenghui, WANG Huiqin+, HU Yan   

  1. School of Information and Control Engineering, Xi'an University of Architecture and Technology, Xi'an 710055, China
  • Online:2018-01-01 Published:2018-01-09

摘要: 针对火焰与背景对比度不明显情况下的低对比度火焰目标提取问题,提出了一种Retinex和CV (Chan-Vese)模型相结合的火焰图像增强和分割算法。首先在YCbCr颜色空间利用Retinex算法构造彩色双边滤波器,根据分量CrCb的关系设计对比度调节函数调节像素点亮度,以凸显原图像中火焰明亮、鲜艳的颜色特征,细节信息也更清晰;再通过帧间差分法和建立的火焰颜色模型获取疑似火焰区域,根据该区域所得中心坐标点设置CV模型的初始轮廓曲线,进一步分割得到火焰目标。仿真实验表明:所提算法不但能够提取出简单和复杂背景环境下低对比度火焰图像,而且目标边缘不规则信息保留完整,误分率比已有算法有明显降低,表明了算法的先进性和有效性。

关键词: 低对比度, 图像增强, 双边滤波器, 图像分割, CV模型

Abstract: For the contrast between fire and background is not obvious, it is hard to extract the low contrast fire. So this paper proposes a fire enhancement and segmentation algorithm based on Retinex and CV (Chan-Vese) model to solve it. Firstly, this paper constructs a color-bilateral filter based on Retinex in the YCbCr color space, and designs a contrast adjustment function to adjust the intensity of each pixel according to the relation of Cr and Cb, which highlights the bright and colorful fire in the image. Secondly, this paper gets the candidate fire area by combining difference frames and color space model, sets the initial contour curve of the CV model according to the center in the area, and obtains the fire area by the further segmentation. The experimental results show that the proposed algorithm can not only extract the low contrast fire in simple and complex background, but also keep the irregular target edge complete, the classification error rate is lower obviously than the existing algorithms. The proposed algorithm is proven to be advanced and effective.

Key words: low contrast, image enhancement, bilateral filter, image segmentation, CV model