Object Identification from Hazed Images, Improving the Boundaries and Corners using KOA Techniques

Abstract

Mist, smoke, and fog will cause objects or images taken from far-off places to be of poor quality and invisible. Here, we use the Optimal Adaptive method (KOA) and K-Means Clustering to rub out or eliminate the undesirable hazes from the image. The photos are first grouped by color clustering using K-Means Clustering (KMC). Our approach greatly benefits from an analysis of the transmission function's intrinsic boundary limitation. To forecast the transmission of the unknown scene, this constraint is treated as an optimization problem. Furthermore, variable splitting methodology-based transmission refinement strategies are employed to address the problem. The suggested approach may recover haze-free, high-quality photos with original colors and improved image information with just a few general assumptions, such as the "K" value and "ω." The results of the experiment, which used evaluation settings and produced 88.07% for our suggested method, show the efficiency and efficacy of the suggested technique with a large number of hazy photos. The suggested work produces better results in FSIM and SSIM when compared to the results that are already available.