People with Colour Vision Deficiencies (CVDs) face notable difficulties in our society that uses colours as a tool of communication in various situations related to design, architecture, traffic, education, etc. Daltonization recolouring tools are a popular strategy in image processing to improve colour perception of people with CVDs by increasing chroma and lightness contrast between confusion colours that are difficult to discriminate for people with CVDs. However, recolouring tools often fall short in practical applicability due to not taking into account basic requirements of various colour tasks, and an insufficient assessment by real people with CVDs. In this paper, we provide guidelines for the design and evaluation of Daltonization recolouring tools to increase practicability and enable their comparison with each other. Namely, a good recolouring tool for people with CVDs (i) should preserve naturalness and originality where possible; (ii) should preserve good colour identification and/or connoted meanings of colours. (iii) should sustain colour communicability consistently throughout the workflow; (iv) should enable customization for different types and severities of CVD of individual users (i.e., it should be open for the integration of different models of the human visual system (HVS)); (v) should define the visual goal of the recolouring tool; (vi) should name the target image type(s) of the tool, e.g. photographs, information graphics, maps, charts; (vii) should account for general restrictions of the medium both in acquisition and reproduction, and should acknowledge challenges related to colour management; (viii) must be tested using real observers with CVDs; and (ix) must be tested on different types of images.
Color Vision Deficiency (CVD) leads to a reduced capability to identify chromatic edges and contrast and may cause significant problems in various color tasks like, for example, comparative color tasks. Many daltonization methods to improve color perception of color-deficient people, however, change naturalness of confusion colors, which might complicate other color tasks like, for example, connotative and denotative color tasks. Thus, we present a daltonization method focusing on the enhancement of chromatic edges and contrast while preserving the naturalness of object colors as good as possible. Our proposed method, Yoshi-II-edge, is based on a previously presented method, Yoshi-II, which projects and rotates the lost information by color-deficient observers along the direction of optimal visibility. Yoshi-II-edge limits this enhancement to chromatic edges and contrast by computing an edge map obtained from the gradient of the error image between the original and its simulation. Furthermore, we propose a threshold and dilation to influence the width of the daltonized edge. We show that the performance of this method depends on the juxtaposition of confusion colors in the image. More precisely, Yoshi-II-edge performs well in images with adjacent areas of confusion colors but performs poorly in images with non-adjacent areas of confusion colors.