Abstract
Affected by the uneven concentration of coal dust and low illumination, most of the images captured in the top-coal caving face have low definition, high haze and serious noise. In order to improve the visual effect of underground images captured in the top-coal caving face, a novel single-channel Retinex dedusting algorithm with frequency domain prior information is proposed to solve the problem that Retinex defogging algorithm cannot effectively defog and denoise, simultaneously, while preserving image details. Our work is inspired by the simple and intuitive observation that the low frequency component of dust-free image will be amplified in the symmetrical spectrum after adding dusts. A single-channel multiscale Retinex algorithm with color restoration (MSRCR) in YIQ space is proposed to restore the foggy approximate component in wavelet domain. After that the multiscale convolution enhancement and fast non-local means (FNLM) filter are used to minimize noise of detail components while retaining sufficient details. Finally, a dust-free image is reconstructed to the spatial domain and the color is restored by white balance. By comparing with the state-of-the-art image dedusting and defogging algorithms, the experimental results have shown that the proposed algorithm has higher contrast and visibility in both subjective and objective analysis while retaining sufficient details.
Highlights
Along with the rapid development of intelligent mining technology integrating the coal industry with big data, image and video technology have come to be widely used in all aspects of mining
We proposed a novel single-channel Retinex algorithm with frequency domain prior information for image dedusting at the top-coal caving face
The method is based on the assumption that the smog-like weather is typically distributed in the lowfrequency symmetric spectrum, while noise and image details are distributed in the highfrequency symmetric spectrum
Summary
Along with the rapid development of intelligent mining technology integrating the coal industry with big data, image and video technology have come to be widely used in all aspects of mining. Retinex algorithms have the characteristic of improving image brightness, which makes them ideal for effective image enhancement and defogging [22] Use of these algorithms may be a suitable and effective method for dust removal and enhancement in a low-light environment such as that of the top-coal caving face. They cannot effectively defog and denoise an image, simultaneously, while preserving the image’s details. We propose a joint dedusting and enhancement method via a singlechannel multiscale Retinex algorithm with color restoration (SC_MSRCR) with frequency domain prior information, to simultaneously remove dust and enhance the details and contrast.
Sign-in/Register to view highlights & summary 
Published Version (
Free)
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call