Fractional Order Approach for Edge Detection of Low Contrast Images

Abstract

In digital image processing, image enhancement techniques are considered important in many applications where the subjective quality of images is important for human interpretation. It serves as a fundamental task to improve interpretability and appearance of an image and is applicable in every field where images are to be understood and analyzed. Contrast enhancement is one of the important enhancement operations in any subjective evaluation of image quality. Many images like underwater images, satellite images, medical images as well as various real time images may suffer from poor contrast due to various factors affecting the surrounding environment during image capturing. Underwater images usually suffer from degraded visibility. Light attenuates and scatters in water resulting in low contrast and haziness in the scenes. Therefore, the main problems to be dealt with in underwater environment are poor contrast, non-uniform lighting, haziness and blurring. Hence, in order to study underwater images, it becomes utmost important to extract the invisible or unclear edges. Since edge detection is widely used in high level processing fields like computer vision, feature extraction, image segmentation etc., various mathematical tools have been developed which aim at identifying these edges in an image. It was indicated that the edge detection methods operationally are a mixture of image smoothing and image differentiation. These integer order differential operators suffer from poor accuracy and noise immunity. In the presented work, a method for edge detection by using fractional order differentiation based approach has been realized. Considering the G-L based fractional differential operator’s basic definition and implementation, a filter is devised and its applicability for texture enhancement is analyzed. Various different underwater images have been used for experimentation using both conventional as well as fractional order differential operators. Further, the results are compared with another approach based on Riemann Liouville (R-L) fractional differential operator. The analysis of tests proves that the proposed method displays better results for detecting edges of low contrast underwater images with high accuracy, good sharpness and reveals more information than conventional methods as well as R-L definition based method.