Gray level image processing using contrast enhancement and watershed segmentation with quantitative evaluation

Abstract

Both image enhancement and image segmentation are most practical approaches among virtually all automated image recognition systems. Feature extraction and recognition have numerous applications on telecommunication, weather forecasting, environment exploration and medical diagnosis. The adaptive image contrast stretching is a typical image enhancement approach and watershed segmentation is a typical image segmentation approach. Under conditions of an improper or disturbed illumination, the adaptive contrast stretching should be conducted, which adapts to intensity distributions. Watershed segmentation is a feasible approach to separate different objects automatically, where watershed lines separate the catchment basins. The erosion and dilation operations are essential procedures involved in watershed segmentation. To avoid over-segmentation, the markers for foreground and background can be selected accordingly. Quantitative measures (gray level energy, discrete entropy, relative entropy and mutual information) are proposed to evaluate the actual improvement via two techniques. These methodologies can be easily expanded to many other image processing approaches.