4.4 Article

Implementation of a Novel, Fast and Efficient Image De-Hazing Algorithm on Embedded Hardware Platforms

期刊

CIRCUITS SYSTEMS AND SIGNAL PROCESSING
卷 40, 期 3, 页码 1278-1294

出版社

SPRINGER BIRKHAUSER
DOI: 10.1007/s00034-020-01517-4

关键词

Field programmable gate array (FPGA); Digital signal processing (DSP) Processor; Zynq; Image de-hazing; Hardware descriptive language (HDL)

资金

  1. Science and Engineering Research Board (SERB) India [EEQ/2016/000556]

向作者/读者索取更多资源

This paper presents a pixel-wise and gray image-based image de-hazing algorithm and compares the performance on different hardware platforms, showing that the Zynq-706 hardware implementation is 1.33 times faster.
Improving the visibility of hazy images is desirable for robot navigation, security surveillance, and other computer vision applications. The presence of fog significantly damages the quality of the captured image, which does not only affect the reliability of the surveillance system but also produce potential danger. Therefore, developing as well as implementing a simple and efficient image de-hazing algorithm is essential. The reconfigurable computing devices like Field Programmable Gate Array and Digital Signal Processing (DSP) processors are used to implement these image processing applications. Several strategies are available for configuring these reconfigurable devices. In this paper, two approaches for hardware implementation of image de-hazing algorithm are presented. The pixel wise and gray image-based de-hazing algorithm is proposed in this paper. The key advantage of this proposed method is to estimate accurate transmission map. It eliminates the computationally complex step of refine transmission map as well as halos & artifacts in the recovered image and achieves faster execution without noticeable degradation of the quality of the de-hazed image. The proposed method is initially verified in MATLAB and compared with the existing four state-of-art methods. This algorithm is implemented on two different hardware platforms, i.e., DSP Processor (TMS320C6748) with floating pointing operations and Zynq-706 fixed-point operations. The performance comparison of hardware architectures is made with respect to Average Contrast of the Output Image, Mean Square Error, Peak Signal to Noise Ratio, Percentage of Haze Improvement and Structural Similarity Index (SSIM). The results obtained show that Zynq-706-based hardware implementation processing speed is 1.33 times faster when compared to DSP processor-based implementation for an image dimensions of 256 x 256.

作者

我是这篇论文的作者
点击您的名字以认领此论文并将其添加到您的个人资料中。

评论

主要评分

4.4
评分不足

次要评分

新颖性
-
重要性
-
科学严谨性
-
评价这篇论文

推荐

暂无数据
暂无数据