Abstract
Purpose: :
Measure the effect of JPEG image enhancement on reaction time (RT) and error rate (ER) in a visual search task.
Methods: :
Subjects pointed to an object in a natural image that match a given search object as soon as possible on a touch screen computer monitor. Three categories of natural images: human group photos (Faces), indoor scenes (Indoor) and collections of similar objects (Collection) were collected from the Internet. Images were randomly presented with or without enhancement. The enhancement was realized by boosting a mid-to-high frequency band in the discrete cosine transform (DCT) coefficients of the image luminance component. Five patients (VA: 20/50 to 20/200) participated.
Results: :
Average RT reduced significantly (p=0.03, Wilcoxon Signed Rank test) with enhancement for both Faces (from 12.2s ± 5 to 9s ± 7) and Collection (from 6.5s ± 2 to 5.8s ± 2). However a non-significant (p=0.3) increase in average RT was observed for Indoor images with enhancement (from 7.1s ± 6 to 8.3s ± 4). The overall average ER with and without enhancement was 19.6% and 23% respectively. The ER was not significantly different between with and without enhancement for all the three categories (p>0.08).
Conclusions: :
Previous studies (Fullerton et al., JOSA A 2007) have shown that patients with impaired vision prefer MPEG enhanced videos. Our results demonstrated an improvement in search task performance with similar image enhancement. Performance improvement was not seen with Indoor images. It is possible that Face and Collection images have to be performed using bottom-up image information whereas for Indoor images, some level of top-down knowledge can be used to aid the search, which might somewhat mask the enhancement effect.
Keywords: low vision • visual search • image processing