Towards a system for segmentation under noisy conditions

This paper reports on an image segmentation system primarily developed for detecting loins in ultrasonic images of live pigs. The images have a low contrast, high level of noise, and a high degree of variance in terms of texture and shape. Our segmentation algorithm starts with a region growing process, which provides a rough approximation of the loin region. Morphological operations and curve fitting eliminate unwanted noise. Finally, an active contour process refines the shape of the resulting region. This method does not rely on specific a priori information of the texture or the contrast. The system is based on the principle of modular design, so that different region growing and refinement algorithms can be easily substituted into the current design as modules. Therefore, the system is general enough to be adapted to other segmentation tasks involving low contrast images.