Abstract
Purpose :
Virtually structured detection (VSD) is an easy, low-cost and phase-artifact free strategy to achieve super-resolution in scanning laser microscopy (SLM). However, in vivo application of this method for high speed imaging of retinal structures is still challenging because of inevitable noises due to illumination power limit allowed for safety concern. The purpose of this study is to explore denoising techniques to enhance the quality of the super-resolution image.
Methods :
Freshly enucleated frog (Rana Pipiens) retina and standard optical target (USAF 1951 1X, EDMOND) were employed to examine the effect of noise reduction in VSD based super-resolution SLM. The super-resolution SLM system employed a 5X objective lens with numeric aperture (NA) 0.1. SLM images were captured with a 300 μW illumination power first. In order to demonstrate the noise effect on the SLM image, the power of the illumination light was reduced to 0.15 μW to get the nosier images. The stacks of 2D light profiles (raw images) obtained from SLM were denoised using a Wiener filter followed by a median filter, and then the noise suppressed 2D light profiles were used to reconstruct a super-resolution image using VSD technique. The super-resolution images with and without denoising involvements were quantitatively compared by calculating the noise level (i.e. standard deviation of noise).
Results :
Figure 1a shows frog retinal image reconstructed with VSD and 1b shows denoising followed by VSD techniques. The results show that the use of noise reduction methods before VSD improved visual quality of the SLM image. The noise level reduced from 0.56 to 0.45 by implementing denoising process before the VSD reconstruction. The denoising effect is more prominent when the low illumination power was used for imaging acquisition. The noise level decreased from 8.70 to 2.48 by adding noise reduction techniques before VSD reconstruction in standard optical target.
Conclusions :
Our experiments show noticeable improvement in the quality of super-resolutions images by implementing denoising techniques in the data processing step before the VSD reconstruction. Further improvement of the VSD method and denoising technique may provide a practical approach to achieve in vivo super-resolution imaging of retinal structures.
This is an abstract that was submitted for the 2016 ARVO Annual Meeting, held in Seattle, Wash., May 1-5, 2016.