Segmentaton Performance Comparison over HDR Images

Abstract

Achievement of sharp, crisp and contrasty photos plays an important role in automatic image segmentation. A fair contrast can be achieved by successful exposure settings in the camera, which can be difficult to achieve during the photo shoot operation. This works aims to clarify effects of post processing single shots (via contrast stretching, which is sometimes called the approximate high dynamic range - HDR) to segmentation performance. It is a known fact that contrast manipulations do not add extra information to the image. On the other hand, most standard segmentation algorithms use fixed threshold settings or fixed methods for threshold selection. Therefore, the contrast processing does have eminent effects on segmentation. This paper presents experimental results on subjective segmentation performances.