Statisticians, as well as machine learning and computer vision experts, have been studying image reconstitution through denoising different domains of photography, such as textual documentation, tomographic, astronomical, and low-light photography. In this paper, we apply common inferential kernel filters in the R and python languages, as well as Adobe Lightroom's denoise filter, and compare their effectiveness in removing noise from JPEG images. We ran standard benchmark tests to evaluate each method's effectiveness for removing noise. In doing so, we also surveyed students at Elon University about their opinion of a single filtered photo from a collection of photos processed by the various filter methods. Many scientists believe that noise filters cause blurring and image quality loss so we analyzed whether or not people felt as though denoising causes any quality loss as compared to their noiseless images. Individuals assigned scores indicating the image quality of a denoised photo compared to its noiseless counterpart on a 1 to 10 scale. Survey scores are compared across filters to evaluate whether there were significant differences in image quality scores received. Benchmark scores were compared to the visual perception scores. Then, an analysis of covariance test was run to identify whether or not survey training scores explained any unplanned variation in visual scores assigned by students across the filter methods.
We employ several Spatial Domain linear filters on grayscale Gaussian noise disrupted images. In addition, we use Adobe's Lightroom denoise method in two forms (50% and 100%) on the same images. Then, benchmark scores are calculated by comparing noiseless photos to the filtered images produced. We were interested in what people thought about the resultant filtered images in comparison to the true noiseless images. Interest in this came from previous research questioning the ability for current benchmark scores to analyze filtering methods. Because image denoising is an NP-Hard problem, only hueristic methods have been produced. This has caused much competition in the field. Evaluating whether benchmark scores are accurately determining methods' abilities to produce improved image quality is important. A shiny app survey was built and 100 Elon University students from introductory mathematics and statistics courses were surveyed under IRB approval. After data collection, an ANOVA test was conducted and then, an ANCOVA test was run using training data from the survey.
The repository is defined as follows:
filtered_images- a folder containing .rdata files of filtered image matricesImage_Quality_Survey- a folder containing the Shiny App survey producedpy_scripts- a folder containing all of the OpenCV methods implemented (Non-local means, bilateral, and 3x3 mean filters)r_scripts- a folder containing scripts for benchmarks scoring, explorative analysis, ANOVA testing, and graph producingsubmitted_paper_draper_taylor.pdf- the final paper submission to the NC Journal of Math and Statssts499_syllabus.pdf- a copy of the syllabus used for grading this course at Elon Universitysas_ancova.txt- a SAS script to run the ANCOVA test