Recent Advances in Corneal Imaging

Abstract

This chapter details work with sequences of corneal images from a confocal microscope to develop enhancement methods to improve the visual quality of the images. Due to involuntary movements of the subject’s eye during image capture, the images suffer both lateral and longitudinal translations, and work is ongoing to attempt to register adjacent images in the sequence. Currently this registration uses an approach based on the Scale Invariant Feature Transforms (SIFT) algorithm. Registration is a necessary stage in the construction of a 3D model of the subject’s cornea for use as a diagnostic aid. The algorithms, results, progress and suggestions for future work are presented in this chapter.

Preparation of 2D sequences of corneal images for 3D model building

“…For 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 7 comparison the Scale Invariant Feature Transform (SIFT) was also applied in a simailar way. This type of registration method (SIFT) works well when the two images are from the same depth but shifted [6].…”

“…9. Two samples showing correct and incorrect matching point found when using original algorithm SIFT [6] To filter the results to remove errors due to spuriously matched points and duplicate matched points as shown in Fig. 9 and get accurate matching between each pair of images as shown in Fig.…”

“…To the best of our knowledge, the analysis of these images are currently based on manual inspection or aided by semi-automatic methods. Individual corneal images in these sequences are often noisy and sometimes dark or contain no data [6,7]. Images display nonuniform illumination caused by factors such as: the spherical shape of the corneal layers, which causes non-uniform reflection of the illumination light from the different areas of the cornea, and the different attenuation of light along the different paths of illumination.…”

“…During a scan, the instrument locates the rear of the cornea (no signal back from the vitreous humour) and then steps forward in 5 µm steps until it reaches the front surface of the cornea (no signal back from the tear layer). This cycle is repeated 3 times during a 20 s scan which provides about 350 images [6].…”

An efficient system for preprocessing confocal corneal images for subsequent analysis

“…8 (b). The work presented in [16] used DCT approach and fuzzy enhancement for preprocessing of the corneal images, and the outputs of these two approaches are shown in Fig. 9 (a) and Fig.…”