| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH |
|
|
|
|||||||
|
|||||||||
Electronic Letters to:
|
|
Electronic letters published:
![[Read eLetter]](/icons/misc/sectionDOWN.gif){kind=icon}
Paragraph Text Reading Using a Pixelized Prosthetic Vision Simulator
|
|
|||
|
Jorge Sommerhalder
Send letter to journal:
jorg.r.sommerhalder{at}hcuge.ch Jorge Sommerhalder
|
Paragraph Text Reading Using a Pixelized Prosthetic Vision Simulator: Inaccurate Quotations of Our Previous Studies We have read with interest the paper by Dagnelie et al.1 dealing with simulations of artificial vision for text reading. In this paper, our previous studies investigating similar issues2 have been reported in an inaccurate way. Indeed, such distorted statements might give a wrong impression about the quality of our work. First, contrary to what is stated by Dagnelie et al., our results are fully pertinent for central vision because the experiments reported in our paper were carried out in central vision. Second, Dagnelie et al. asserted that we did not find a difference in reading performance when comparing Gaussian versus square pixelization, because we did not use real-time implementations of the algorithms. This is inaccurate. We did compare the two pixel profiles with both real-time and off-line filters. Off-line filters yielded a slight (but significant) advantage to Gaussian pixelization (experiment 2), but this advantage vanished with the use of real-time filters (experiment 3). Dagnelie et al. also mentioned that in our studies, reading performance dropped for grid densities of 5 pixels/charwidth, which is slightly above the 4 dots/charwidth that they found in their experiments. However, this comparison is inadequate because it is based on static stimuli. For real-time pixelization (experiment 1 in our paper), this threshold pixel density decreased at approximately 3.6 pixels/charwidth, which is entirely consistent with the results reported in their paper. It is, however, worth mentioning that in our experiments subjects achieved reading speeds about twice as fast as theirs. Furthermore Dagnelie et al. suggested that our simulations were not pertinent for future users of retinal prostheses. For example, they say pertinent simulations should present light text on a dark background, because future prostheses will be built this way to avoid "problems of glare and stimulation overload." This might be true. However, this does not represent a realistic situation since most written material is presented as dark text on a light background. We adapted our simulations to this reality. Moreover, prostheses using such an inverse-contrast transformation might be beneficial only for reading, but would not make sense for other tasks. We believe that the results of both papers complement each other. The above-mentioned clarifications seemed to us however necessary so that readers of IOVS can correctly evaluate the results reported by each research team. Angélica Pérez Fornos Ophthalmology Clinic, Geneva University Hospitals, Geneva, Switzerland References 1. Dagnelie G, Barnett D, Humayun MS, Thompson RW Jr.
Paragraph text reading using a pixelized prosthetic vision simulator: parameter dependence and task learning in free-viewing conditions. Invest Ophthalmol Vis Sci. 2006;47:1241-1250. |
|||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH |
eLetters: