In this chapter, different approaches are presented for removal of fog from video footage taken in moving cars. The methodology uses different approaches, namely dark channel prior, contrast limited adaptive histogram equalization (CLAHE), the combination of two approaches (dark channel prior and CLAHE), and RETINEX algorithm combined with DWT. The algorithms are implemented in MATLAB R2015a. Moreover, the algorithms are compared based on their computational complexity and a visibility metric which is used for computing the CNR of video frames before and after the application of the algorithm. The chapter discusses which algorithm would provide better performance during night fog and daylight fog. Finally, the safe speed of the driver is calculated based on the time complexity of the algorithm used.
TopIntroduction
Humanity has witnessed a tremendous increase in population along with large-scale urbanization in the past decade. As the economic status of people continues to improve considerable investment is done in the field of technology applications for enhancing the lives of people. The smart cities are developed with the intent to provide an improved lifestyle to its citizens using data and technology-driven solutions to tackle real-life problems. Various data collection and processing techniques are employed in the public systems to extract useful information and aid in the management of various systems like traffic, law enforcement, schools. The word ‘smart’ in smart city refers to the application of intelligence in the above systems.
Security and safety of people is an important indicator of the development of a city and cannot be neglected. It is not uncommon to witness cases of trains getting derailed, flights getting delayed and collision of vehicles on road due to poor visibility. The bad climate conditions considerably decrease the visibility of the drivers which makes it almost impossible to drive. Owing to the steady rise in population in the cities there has been heavy traffic due to which effective management of Traffic and Transportation Systems becomes important. Fog, mist, rain etc are examples of some natural phenomenon which are a major cause of road casualties. Similarly, in the industrial areas smog (a mixture of smoke and fog) is a commonplace phenomenon which considerably degrades human vision.
The main aim of this chapter is to devise such a system which can provide clear footage of the road to the driver and thus assist in driving. The system is simulated using the MATLAB R2015a graphical user interface.
This chapter starts off with introductory definitions of image and video processing along with the fields associated with them. This is followed by the need for fog removal and literature review of algorithms which have been implemented in the past for fog removal in images. The chapter further implements these algorithms and their combinations on video feeds which have been captured from moving cars
Two major constraints need to be considered for real-time traffic monitoring:
This chapter focuses on three major algorithms namely Dark Channel Prior, CLAHE and Retinex algorithm in the process of fog removal which has been used previously mainly on images A comparative analysis of the techniques provided with respect to time delay and visibility. Visibility is defined in terms of CNR (Contrast to Noise Ratio). The chapter develops and analyses the above-mentioned techniques for fog removal in video feeds. This will give the driver a clear view of the road while driving thereby helping in the prevention of accidents
The chapter begins with the dark channel prior technique. In this technique, the least value for each pixel in the image is calculated using minimum filtering. A model of the foggy image is considered which consists of atmospheric light A and transmission map t. These are estimated for the reconstruction of a fog-free image. In the end, guided filtering is applied to remove residual fog from the video frame.
The Contrast Limited Adaptive Histogram Equalisation (CLAHE) algorithm establishes a maximum value to clip the histogram and redistributes the clipped pixels equally to each gray level. It limits the noise while enhancing the contrast. Firstly, the background image is extracted from the video sequence followed by which the moving pixels are estimated and bounded into foreground images. Secondly, the foreground and background images are defogged respectively by CLAHE. Lastly, the foreground and background images are fused into the new frames.
The retinex algorithm deals with compensation for illumination effects in images. The aim is to decompose a given image S into two different images, the reflectance image denoted by R, and the illumination image denoted by L, such that at each point (x,y) in the image domain S(x,y)=R(x,y).L(x,y). The method implemented here makes use of the Retinex algorithm to enhance the image, then ‘haar' wavelet transform is used to enhance the details of the image. Finally, a clear image which has the fog removed can be obtained after reducing the non-important coefficients.