Computer-Based Solutions to Support Those With Colour Vision Deficiency to Access Day-to-Day Information

Abstract

In modern-day society, we are bombarded with vast amounts of electronic information which we may be expected to make decisions from. Many people have difficulties in interpreting such information due to either physical or cognitive difficulties in using electronic devices, or an inability to identify information as intended by the author.

Colour Vision Deficiency (CVD) is one such problem that can cause considerable difficulty in the interpretation of diagrammatical information. This is because a Colour Vision Deficient (CVDt) person has difficulty in seeing: colour boundaries, different shades of colour and different hues. There has been some research to aid the CVDt, where the majority of the research in image processing changes or transforms colours in any given image. Such transformations use a number of different algorithms to create a CVDt friendly post-processed image from the pre-processed image. A major problem of current transformation algorithms is that they are aimed for specific contexts and cannot be used in generic contexts. For example, the transformation algorithm may be aimed at aiding the CVDt to view postprocessed images of weather maps only.

The aim of this dissertation is to provide an improved post-processed image algorithm. The algorithm is intended to provide the CVDt with greater benefit by being able to interpret the information in the post-processed image correctly. The algorithm used in this dissertation is not a colour transformation algorithm instead it is a colour separation algorithm. This concept of colour separation is novel.

The colour separation algorithm, which is called the Halo-Effect Algorithm (HEA), parses a given image row-by-row and pixel-by-pixel until the end of file-marker is reached and a CVDt friendly post-processed image is furnished. When there is a colour change between two identified pixels then a colour boundary has been identified within the pre-processed image and a differently coloured pixel is inserted between two, furnishing the post-processed image. As the pre-processed image is parsed row-by-row then the colour the boundary builds up to form a colour boundary interface where the different coloured pixel are inserted in the post-processed image. In this dissertation the separation pixel is always white. The build-up of inserted white pixels at the colour boundary interface of the pre-processed image produces a halo like effect in the post-processed image which is CVDt friendly.

To demonstrate the efficacy of the colour separation concept, the HEA has been developed and implemented. A number of surveys have been conducted using participant responses to questions within each survey. The responses that each participant gave were then collated and analysed statistically. Two statistical techniques were used to test a number of hypotheses around the mean of a sample drawn from a normally distributed population. In this dissertation the normally distributed populations were the survey participants. From the analyses of the responses, the survey population was divided into two groups. One group was identified to have no problem with identification of pre-processed colour boundaries and were called the non-CVDt. A second group was identified to be those who had some problems with the identification of pre-processed colour boundaries and were called the indicative- CVDt.

Responses from the two groups were collated and statistical analyses were then conducted to test the significance of any results obtained and also to test the validity of the algorithms under investigation. In this dissertation two currently available, but different, colour transformation algorithms were compared with the colour separation algorithm of the HEA. Each of the two transformation algorithms were originally intended for specific use. One was aimed for spectra maps and the other was aimed for background text. Statistical analyses showed that each of the transformation algorithms provided benefit to the indicative-CVDt for their specific context only. However, statistical analyses also showed that HEA fared well in each of the two specific contexts. Thus, hinting that colour separation of HEA could be used in more general contexts.

To confirm that colour separation can provide greater benefit to the indicative-CVDt in more generic contexts than colour transformations further surveys were undertaken. In each survey participants were asked a number of questions about a given image where colour boundaries are expected to occur frequently. One was a map of the provinces of Australia and the other a number of differently coloured geometric shapes. Statistical analyses showed that the colour separation algorithm of HEA provided greater benefit to the indicative-CVDt than the two colour transformation algorithms in both cases. Hence, confirming that colour separation of HEA is beneficial to the indicative-CVDt in generic contexts.

Colour separation of the HEA is still in its infancy and a great deal more research is required to determine how great its efficacy is. For example, clinical studies could be undertaken using two sets from one population. One set of participants who would have been diagnosed as non-CVDt, which would be identified as a control group, and a second set who would have been diagnosed as CVDt, which would be identified as a test set.