Image dehazing based on dark channel spatial stimuli gradient model and image morphology
- PDF / 5,107,933 Bytes
- 13 Pages / 595.276 x 790.866 pts Page_size
- 102 Downloads / 226 Views
ORIGINAL RESEARCH
Image dehazing based on dark channel spatial stimuli gradient model and image morphology Rehan Mehmood Yousaf1 · Hafiz Adnan Habib2 · Zahid Mehmood2 · Muhammad Bilal1 Received: 2 October 2019 / Accepted: 27 September 2020 © Springer-Verlag GmbH Germany, part of Springer Nature 2020
Abstract Image dehazing has become a critical problem to cater to as it has several parameters that need to be addressed. Real color, contrast, and illumination are the major parameters that are to be restored in the dehazed image. Different scenarios distort these parameters in different ways so it is difficult to restore the original image. Many approaches are used in the literature to cater to these problems but suffer from low contrast, faded color, and weak edges. This article introduces an effective technique, which is named as dark channel spatial stimuli gradient model (DCSSGM) that performs well for the aforementioned problems. The DCSSGM technique applies the dark channel prior (DCP) and spatial stimuli gradient sketch model (SSGSM) on each color channel to eliminate the haze from the image and to restore true edges. SSGSM is responsible to restore robust edges in an image using the perceived brightness and calculations based on neighborhood similarity. Morphology is applied to the resultant image to receive sharp true edges. The final restored image is the output of dynamic histogram equalization (DHE) which restores the contrast of the image. The evaluation analysis qualitatively and quantitatively concludes that the DCSSGM technique outperforms other state-of-the-art image dehazing techniques. Keywords Image dehazing · Edge preservation · Dark channel prior · Spatial stimuli gradient model
1 Introduction The understanding of real-world natural images plays a vital role in computer vision and image processing techniques. The computer visual system tells us the quality of perception in real-world natural images. The quality of perception can be analyzed by the naked eye test, qualitative, and quantitative analysis. Many visual methods are based on the quality of perception such as object detection, recognition, and surveillance. A hazy image is formed due to the presence of particles in the atmosphere that reduces the clarity of the natural scenes. The haze could be fog, smog, or a mixture of multiple including mist. Due to these unwanted particles, the light is dispersed in the image, and the original illumination of the image is disturbed. The color and contrast of
the image are also affected in the hazy image (Koschmieder 1924). Techniques related to image processing and computer vision may need a haze-free image depending upon the nature of the application. As the hazy images contain lots of extra and scattered light, so this becomes problematic for the image processing algorithms. Similarly, the real scene is very significant and needs to be accurate so that computer vision perception is clear and analysis is done accordingly. Keeping these facts into view, the haze has become an issue at a low-le
Data Loading...
