Logistic Regression Model for Image Preference Estimation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for evaluating image and video quality rely on subjective testing, which is uncertain and expensive, and do not effectively account for viewer preferences and distortion types, leading to unreliable quality estimates.
Innovation Solution
A predictive model using logistic regression is developed to estimate the probability of image preference based on viewer data, incorporating image effects, viewer effects, and distortion type effects, allowing for well-calibrated pairwise image preference predictions and confidence levels.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If subjective testing using people is used to evaluate image and video quality, then quality assessment can be performed, but the process becomes expensive and uncertain
Solution Approach 1:
The patent creates a predictive model that copies human subjective quality assessment behavior through logistic regression trained on viewing preference data. Instead of repeatedly conducting expensive subjective tests, the model replicates human judgment patterns to estimate image preference probabilities, maintaining assessment accuracy while eliminating the need for continuous human testing.
Solution Approach 2:
The patent replaces the mechanical system of human subjective testing with an automated computational model. The logistic regression model processes image features and distortion types to predict quality outcomes, substituting human observers with an algorithmic system that provides consistent, scalable, and cost-effective quality assessment.
2Productivity
If absolute quality scores are used to estimate image quality, then a single score is obtained, but the scores are unreliable for comparing viewer preferences
Solution Approach 1:
The patent transitions from one-dimensional absolute quality scores to a two-dimensional probabilistic framework that estimates both the preference outcome and the confidence level. By adding the probability dimension (p > 0.5) and confidence interval, the model provides richer information that enables more reliable preference comparisons while maintaining computational efficiency.
Solution Approach 2:
The patent changes the output parameter from a single deterministic quality score to a probabilistic estimate with confidence levels. This parameter transformation allows the model to express uncertainty and provide calibrated predictions, improving reliability for preference comparison tasks while preserving the efficiency of automated computation.
3Loss of information
If viewer preference data is collected through subjective testing, then preference information is obtained, but the process is time-consuming and expensive
Solution Approach 1:
The patent performs preliminary action by collecting and analyzing viewer preference data upfront to train the logistic regression model. Once trained, the model can rapidly predict preferences for new images without requiring additional subjective testing. This preliminary data collection phase enables subsequent fast, automated preference estimates, reducing overall time and cost.
Solution Approach 2:
The patent implements self-service by having the predictive model generate preference estimates autonomously without requiring ongoing human involvement. The model uses its trained parameters to independently assess new images, eliminating the need for continuous subjective testing while maintaining preference data quality through the robust probabilistic framework.
Data Source
AI summary
A system that incorporates teachings of the present disclosure may include, for example, sampling a variable effect distribution of viewing preference data to determine a first set of effects comprising a plurality of first distortion type effects associated with a first distortion type of a first image and to determine a second set of effects comprising a plurality of second distortion type effects associated with the second distortion type of a second image, calculating a preference estimate from a logistic regression model of the viewing preference data according to the first set of effects and the second set of effects, wherein the preference estimate comprises a probability that the first image is preferred over the second image, and selecting one of the first distortion type or the second distortion type according to the preference estimate. Other embodiments are disclosed.