Asian Journal of Engineering and Technology Innovation

Ghost Detection and Removal in HDR Images

Vidhya. R, Monika. K, Shwetha. M, Swetha. K. N, Pavithra. P

Dept of CSE, REVA ITM, Bangalore

Abstract

Digital cameras can only capture a limited range of images, when a user tends to take a photograph of any scene the areas that have a high intensity looks brighter and suppose if the image is captured from a dark scene the images will be of low intensity that is less brighter. When these images are taken in either of these scenes the output image will be a combination of images with low dynamic range. In order to get a clear and a high quality range of images, the technique was developed to obtain a high quality of image in a more convenient way that is known as high dynamic range imaging. The output images are called high dynamic range (HDR) images, since there were many techniques that were evolved for fusion. The two methods are fusion in radiance domain and fusion in image domain. The main criteria for taking a combination of several images are to take the image that is placed at a particular position that is constant. So, any object that has a motion in the scene it will result in causing a blurred image that is called as ghosting artifacts in the output HDR image. To identify the ghost objects in HDR image there are many methods namely, Variance based ghost detection, Entropy based ghost detection, Prediction based ghost detection, Pixel order relation, one of these techniques are used to identify ghost objects in the image.

Eliminating the ghost object in the final HDR image is the main agenda of any method that solves the ghost problem. Many methods produces a different kind of outputs and can be divided into two main types. One is keeping a object at one particular position which has a motion, and another is eliminating all the objects that has a motion in it.

Keywords:

low quality range, fusion, entropy, ghost identification, ghost objects, ghost elimination.