Presenter(s)

Sai Babu Arigela

Download

Download Project (2.2 MB)

Description

A new automatic image enhancement technique based on a locally tunable transformation function for visibility improvement in aerial images is presented. Aerial images usually suffer with poor visibility and contrast because of bad weather conditions like haze, fog, and turbid conditions. We propose a model based image restoration approach which uses a new nonlinear transfer function on luminance component to obtain the transmission map. The model assumes that the weather conditions include haze and fog particles. The amount of accumulation of haze/fog particles depends on the depth information of the scene. The local luminance image provides approximate depth information of haze/fog regions. Local multi scale Gaussian mean is used to estimate the approximate local depth image. A new nonlinear function which is locally adaptive based on the approximate local depth information is used to estimate the transmission map of the image. The haze free image can be restored from the haze image by estimating the transmission map and substitute in the model for each spectral band. Results from various experiments demonstrate that this technique can be used for various applications like traffic monitoring, weather observation, video surveillance, and security applications.

Publication Date

4-17-2013

Project Designation

Graduate Research

Primary Advisor

Vijayan K. Asari

Primary Advisor's Department

Electrical and Computer Engineering

Keywords

Stander Symposium project

An Automatic and Locally Tunable Transformation Function for Fog and Haze Removal in Aerial Imagery

Share

COinS