|
Diagnosis and treatment of retinal diseases such as diabetic retinopathy commonly rely on a clear view of the retina. High quality retinal images are essential in early detection and more accurate diagnosis of many retinal diseases. Conventional fundus cameras usually lack the ability to provide high resolution details required for diagnostic accuracy. Major factors contributing to the degradation of retinal image quality are the aberrations from the eye and the imaging device. The challenge in obtaining high quality retinal image lies in the design of the imaging system that can reduce the strong aberrations of the human eye. Since the amplitudes of human eye aberrations decrease rapidly as the aberration order goes up, it is more cost-effective to correct low order aberrations with adaptive optical devices while process high order aberrations through image processing. A cost effective fundus imaging device that can capture high quality retinal images with 2-5 times higher resolution than conventional retinal images has been designed. This imager improves image quality by attaching complementary adaptive optical components to a conventional fundus camera. However, images obtained with the high resolution camera are still blurred due to some uncorrected aberrations as well as defocusing resulting from non-isoplanatic effect. Therefore, advanced image restoration algorithms have been employed for further improvement in image quality. In this paper, we use wavefront-based and self-extracted blind deconvolution techniques to restore images captured by the high resolution fundus camera. We demonstrate that through such techniques, pathologies that are critical to retinal disease diagnosis but not clear or not observable in the original image can be observed clearly in the restored images. Image quality evaluation is also used to finalize the development of a cost-effective, fast, and automated diagnostic system that can be used clinically.
|
