Blurred Eye Research Articles

Adam Optimization Method Modifications Comparison in the Regression Models Parameters Evaluation Tasks

There is considered a problem of optimization methods comparing for the neural network regression task. Various optimization methods, such as stochastic gradient descent with momentum (SGDM), Adam and its modifications, AdamW and RAdam, were considered. To compare optimization method two regression task were formulated. Both tasks are connected with the preprocessing subtasks in the field of image analysis. The first considered task was the filtering blurred eye images on which confident recognition cannot be achieved. The training samples were generated by Gaussian blurring of the images. The blurring degree was estimated. The test and training sample for the assessment problem was formed on the basis of the BATH and CASSIA eye image databases. The second task was aligning faces in assessment image in face recognition systems. The training samples were generated by rotating face images, and rotation angle was estimated. To solve these tasks the direct estimation of parameters by solving the image regression problem by training neural network models is proposed. The adequate accuracy was acquired with all considered optimization methods for both tasks. Modifications of Adam algorithm show better results than original method. Both AdamW and RAdam methods reduced the error twice in comparison with Adam. The modification of the RAdam algorithm proposed in the work reduced the error by more than 1.5 times in comparison with the model trained by the original algorithm.

Investigation of interocular blur suppression using luminance-modulated and contrast-modulated noise stimuli.

Presenting two sufficiently dissimilar images, one to each eye, may result in interocular suppression. The present study measured interocular suppression depth and extent in binocularly normal participants when blurring one eye only with varying dioptric lens powers (+0.5, +1, +2, and +4 D). Visual stimuli consisted of eight concentric rings of alternate polarity, divisible into eight sectors, within the central circular 24° visual field. Binocular "ring" stimuli therefore consisted of 64 individually testable dichoptic sectors. Using a two-alternative forced choice paradigm with a staircase procedure, signal strength of each dichoptic sector in the blurred eye was adjusted to perceptually match that of the surrounding ring from the nonblurred eye, determining the point of subjective equality. Rings were defined by differences in luminance (L), luminance-modulated noise (LM), or contrast-modulated noise (CM). Suppression depth was similar irrespective of sector location within the visual field and increased with increasing difference in interocular blur. Adding dynamic noise (LM vs. L stimuli) reduced the effect of blur on measured suppression depth. Significantly deeper suppression was measured for CM than for LM stimuli, both created using dynamic noise, the difference increasing at higher levels of interocular blur. As binocularity is disrupted with interocular blur, this result suggests that CM envelope combination may be processed by later mechanisms receiving binocular input than those required for the processing of LM stimuli. Differences in suppression depth between LM and CM stimuli could not be attributed to differences in spatial summation properties, stimulus visibility, noise modulation, or differential effects on blur discriminability.

Dichoptic Suppression of mfVEP Amplitude: Effect of Retinal Eccentricity and Simulated Unilateral Visual Impairment

To investigate the effect of retinal eccentricity on the phenomenon of dichoptic suppression of the mfVEP amplitude and to examine the relationship between the degree of simulated unilateral visual impairment and the possible release of dichoptic suppression in the contralateral eye. Eight subjects with corrected visual acuity (VAc) >6/6 and stereoacuity >60 sec arc underwent monocular and dichoptic pattern-pulse mfVEP. Dichoptic stimulation was repeated with refractively induced blur of one eye with +4-D and +6-D lenses above distance correction. Dichoptic recording resulted in significant reduction of averaged mfVEP amplitude (19.8% ± 4.9%, paired t-test, P = 0.00003). The magnitude of suppression, while statistically significant at all eccentricities, was significantly larger in the central part of the visual field and diminished toward the periphery. Refractive blur, used to simulate visual impairment produced variable degrees of amplitude reduction in the blurred eye and resulted in amplitude increases in the contralateral eye. There was a highly significant correlation between the magnitude of amplitude reduction in the blurred eye and increase in amplitude (i.e., release of dichoptic suppression) in the contralateral eye (r = 0.91, P < 0.0001). The study demonstrated that dichoptic stimulation results in eccentricity-dependent suppression of mfVEP amplitude. Factors affecting visual performance of one eye (monocular blur) promote the release of dichoptic suppression in the fellow (unaffected) eye. This phenomenon leads to an increase in intereye asymmetry and therefore may improve early detection of ocular diseases, especially monocular pathologic processes.

An electrophysiological correlate of perceptual suppression in anisometropia

Cortical evoked potentials (EP) in response to phase-alternating gratings were recorded in subjects with various degrees of artificial anisometropia. obtained by inserting convergent lenses in front of one eye. The EP recorded during binocular viewing with artificial anisometropia failed to show an increased amplitude in comparison to the monocular EP, and the lack of “binocular summation” of the EP was found to be related to perceptual suppression of the blurred eye. These results provide a method for testing monocular suppression in patients with anisometropia or other defects of binocular vision.