This repository contains code for analyzing the relationship between scene image properties and perceptual judgments
Start by navigating to the scene-filter-see "DoAnalysis" directory in MATLAB. The script "do_analysis" is a good place to start. You can run it with a few different analyses, for instance do_analysis('image_correlation') to perform an image-based analysis of the predictiveness of whole-image luminance/depth correlations on perceived 3D.
Other folders:
IMAGE MANIPULATION: high-res versions of the original and manipulated images used in the experiments; Matlab code that generated the manipulations, which can be altered and run within the current directory
PERCEPTUAL EXPERIMENT: response data from 20 subjects from two perceptual experiments
Each .mat file contains a structure with the data from the main experiment or the control experiment.
There are 3 upper level fields:
-conditionTypes: the image manipulations being compared -'tp' is towards the prior -'ap' is against the prior -'orig' is unmanipulated -'stereo' is stereo 3D
-scenesListSorted: directory names for the 15 scenes used in the experiment, sorted from most negative to most positive luminance/depth correlation. These file names match the directories in ImageManipulation
-scenesListSortedRvals: Pearson r correlation values between luminance and depth for these scenes
-trials: contains 7 field with a row for each trial in the experiment -subj %subject number -scene_number %index of original scene - can be used with scenesListSorted to determine scene name -condA %one condition from trial - indexes into conditionTypes to determine the manipulation -condB %other condition from trial - indexes into conditionTypes to determine the manipulation
-AcorrMinusBcorr %did condition A's image have a more negative correlation than B?
-AvarMinusBvar %did condition A's image have greater image variance than condition B?
Responses:
for main experiment:
-resp_Amore3DthanB %was condition A judged more 3D than B?
for control experiment:
-resp_AmoreContrastthanB %was condition A judged more Contrasty than B?