Enhanced Variational Image Dehazing

Abstract

Abstract. Images obtained under adverse weather conditions, such as haze or fog, typically exhibit low contrast and faded colors, which may severely limit the visibility within the scene. Unveiling the image structure under the haze layer and recovering vivid colors out of a single image remains a challenging task, since the degradation is depth-dependent and conventional methods are unable to overcome this problem. In this work, we extend a well-known perception-inspired variational framework for single image dehazing. Two main improvements are proposed. First, we replace the value used by the framework for the grey-world hypothesis by an estimation of the mean of the clean image. Second, we add a set of new terms to the energy functional for maximizing the inter-channel contrast. Experimental results show that the proposed Enhanced Variational Image Dehazing (EVID) method outperforms other state-of-the-art methods both qualitatively and quantitatively. In particular, when the illuminant is uneven, our EVID method is the only one that recovers realistic colors, avoiding the appearance of strong chromatic artifacts.

Fusion-based Variational Image Dehazing

“…To restore haze-degraded outdoor images, authors of [22] proposed to minimize the following image energy:…”

“…To minimize the E FVID energy, the Euler-Lagrange equations for E FVID can be easily derived (see [22]):…”

“…Another seldom addressed problem is the appearance of chromatic artifacts in sky regions [27], [28], Here, we take [4], (c) NBP [3], (d) BCCRD [6], (e) EVID [22], and (f) FVID. (g) SSIM of partial images produced masking dehazed outputs by binarized depth maps: (h)-(k) show progressive masked images at a depth of: (h) 4% (i) 16% (j) 36% (k) 64%.…”

“…(g) SSIM of partial images produced masking dehazed outputs by binarized depth maps: (h)-(k) show progressive masked images at a depth of: (h) 4% (i) 16% (j) 36% (k) 64%. [5], (e) EVID [22], and (f) FVID.…”

“…To that end, we design a three-step procedure. We first minimize the EVID energy, which was designed for image dehazing [22], but tends to generate some overenhancement on nearby areas. Instead of retaining only the last EVID iterate as the dehazed output, we store images associated to all iterates, corresponding to progressively dehazed images.…”