Exposure versus contrast enhancement in low-light 4k videos
- PDF / 1,908,130 Bytes
- 8 Pages / 595.276 x 790.866 pts Page_size
- 48 Downloads / 247 Views
ORIGINAL PAPER
Exposure versus contrast enhancement in low-light 4k videos Roha Ali1 · Adeel M. Syed1 · Syed Sohaib Ali2
· Muhammad Mohsin Riaz3 · Muhammad Zafar Iqbal4
Received: 31 October 2019 / Revised: 7 February 2020 / Accepted: 19 May 2020 © Springer-Verlag London Ltd., part of Springer Nature 2020
Abstract The terms “exposure enhancement” and “contrast enhancement” are usually used interchangeably for image enhancement in low-light acquisitions. This paper argues that the two processes are anything but alike. While exposure can be inferred from low frequency or background of an image, contrast by definition refers to the ratio of the foreground with the background. We break down the Lambertian illumination model in a multiresolution framework and show how foreground and background at different levels contribute to contrast and exposure enhancement. Following this, a fast multiresolution low-light enhancement scheme is proposed that achieves a frame rate of 19.62 frames per second on standard quad core processor without employing any graphical processing unit. Experiments reveal that the proposed scheme is not only fast, it also achieves state-of-the-art performance in terms of subjective and objective assessment. Keywords Contrast enhancement · Exposure correction · Low-light video enhancement
1 Introduction Images acquired by visual sensors during night are under exposed due to limitations in sensor’s sensitivity to visible light, shutter speed of the camera and the lens aperture. This exposure can be corrected locally after image acquisition but most of the schemes devised for this purpose are computationally expensive and thus unsuitable for real-time video enhancement. With modern cameras providing 4K resolution
B
Syed Sohaib Ali [email protected] Roha Ali [email protected] Adeel M. Syed [email protected] Muhammad Mohsin Riaz [email protected] Muhammad Zafar Iqbal [email protected]
1
Software Engineering Department, Bahria University Islamabad, Islamabad, Pakistan
2
Computer Science Department, Comsats University Islamabad, Islamabad, Pakistan
3
Center for Advance Studies in Telecommunication, Comsats University Islamabad, Islamabad, Pakistan
4
Technical University of Munich, Munich, Germany
of 3840 × 2160 pixels @ 30–90 fps, a fast contrast enhancement technique is also imperative. The term “contrast” has a variety of interpretations in optics. The most simplest one can be interpreted as the ratio between luminance difference and average luminance. This ratio, formulated by Ernst Heinrich Weber [7], is also called just noticeable difference (JND) with which an object with can be distinguished relative to the background, i.e., JND =
I I
(1)
where I represents the smallest increment or decrement in the object’s intensity in a relatively constant background. The assumption here is that the object is relatively small compared to the background so that the average luminance is approximately equal to the uniform background. This assumption of a uniform backgrou
Data Loading...
