Multi-scale retinex-based contrast enhancement method for preserving the naturalness of color image
- PDF / 2,626,886 Bytes
- 11 Pages / 595.276 x 790.866 pts Page_size
- 38 Downloads / 188 Views
REGULAR PAPER
Multi‑scale retinex‑based contrast enhancement method for preserving the naturalness of color image Shi Bao1 · Shaoying Ma1 · Chuanying Yang1 Received: 6 June 2020 / Accepted: 20 August 2020 © The Optical Society of Japan 2020
Abstract For images with insufficient visibility, image processing is required. Retinex theory is often implemented for image contrast enhancement for images with this characteristic. This paper proposes a multi-scale retinex-based contrast enhancement method using the illumination component; this method could preserve the naturalness of color images. In the proposed method, illumination was first modified using an illumination modification factor; second, the image was enhanced via adaptive gamma correction. Finally, through the combination of the illumination components of the input image and the adaptive gamma correction image, we ensured the visibility and the naturalness of the output image. To confirm the effectiveness of the proposed method, we compared it with existing contrast enhancement methods. For the experiments, we employed discrete entropy, lightness order error, and mean chrominance error to perform the numerical evaluation. The results indicated that our method was better than a majority of existing methods. Moreover, with regard to the visual evaluation, the naturalness of the image obtained via the proposed method was superior to that of images obtained using other methods. Keywords Contrast enhancement · Hue · Saturation · Value (HSV)color space · Illumination component · Retinex
1 Introduction When capturing an image with a device, any type of disturbance may lead to deterioration in the quality of the image. In low-light conditions, objects in captured photos often feature insufficient (or poor) visibility. The poor performance of the camera or the insufficient professional knowledge of the camera operator could account for this phenomenon. For example, the objects details become blurred, resulting in a reduction in image quality. These issues can cause adverse effects on vision tasks such as object detection, recognition, and tracking. Techniques called image enhancement techniques are employed to deal with such images; these techniques can * Shaoying Ma [email protected] Shi Bao [email protected] Chuanying Yang [email protected] 1
College of Information Engineering, Inner Mongolia University of Technology, 49 Aimin Street, Xincheng district, Hohhot 010051, China
improve the visibility of objects. Previous studies have proposed numerous image enhancement methods, mainly focusing on two domains: the spatial domain [1–10] and the transform domain [11–20]. The most common spatialdomain based image enhancement methods are the histogram equalization (HE) [1] and its improved versions. For example, Huang et al. proposed a contrast enhancement method utilizing the adaptive gamma correction with weighting distribution (AGCWD) [10]. In the AGCWD, the probability distribution of luminance pixels is obtained and gamma correction is subsequently performed. Moreove
Data Loading...
