@phdthesis{oai:mie-u.repo.nii.ac.jp:00012627,
 author = {Pakizar, Shamoi},
 month = {Mar},
 note = {application/pdf, Current dissertation proposes the novel approach for color information representation and processing using fuzzy sets and logic theory. Our method is based on the fuzzification of the well-known HSI color space. Specifically, we use fuzzy mathematics to partition the gamut of feasible colors in HSI space based on standard linguistic tags. As a result we obtain a set of fuzzy colors, a human-consistent color model - FHSI (Fuzzy HSI) color model. In FHSI color channels distributions are expressed with fuzzy membership functions. Soft boundaries between color zones were defined empirically. In fact, membership functions were derived from the experimental results we got from a survey. In FHSI, colors are modeled considering the imprecision, subjectivity, context dependency and non-uniformity of color distributions.
   Fuzzy sets are very suitable for this purpose since humans have different levels of visual sensitivity and different color perception abilities. Fuzzy logic is tolerant of imprecise data, so it enables us to make borders between red and orange, harmonious and nonharmonious somewhat blurred. In addition, fuzzy approach allows us to define query conditions on the basis of linguistic terms, which is more natural way for a user to express his desire. It also allows to account for the non-uniformity of color distributions. This method enables to directly model colors such ""light blue"" or ""deep red"" and retrieve images based on fuzzy dominant colors expressed through linguistic descriptions.
   We also provided objective measures for finding the image similarity in a way that matches human evaluation. Our formula takes into account the notion that different hues have various value ranges, due to linguistic conventions of the society (e.g. green color). In our color difference method, we use the saturation as a weighting factor, giving priority to the hue when the saturation is high, and giving priority to the intensity value when colors have low saturation.
   Colors are in general imprecise. For example, it is not possible to draw a clear boundary delimiting blue and green colors. So, ""blue"" is a gradual concept and the boundary between what is blue and what is not blue is fuzzy. By adopting fuzzy set representations we solve the problem of the semantic gap between color representation in computers and high-level linguistic terms and aesthetic concepts. By performing color feature extraction of images, FHSI model is further exploit in the application of image indexing and retrieval.
   The currently most popular and widely used approach for image retrieval is based on text annotations. Thus, images are associated and indexed with certain keywords and tags, which are used to search for images. However, TBIR has limitations like subjectivity (i.e., different people may interpret an image differently) and incompleteness because they do not necessarily reflect the low-level image features very well. Moreover, TBIR systems require humans to manually describe every image in the database, which is obviously impractical for large databases.
   Developed apparel online shop with underlying fuzzy color processing mechanisms allows to overcome the limitations of the traditional e-commerce search by providing automatic labelling. It supports the processing of three types of queries: linguistic, exemplar(involving a reference image) and combinational.
   Our system has two important parts: assigning a fuzzy colorimetric profile (indexing stage) to the image and processing the user query (retrieval stage). System is also able to process impressions like formal (dark colors), creative (yellow, orange), luxury(violet,golden), romantic, warm, pastel, elegant, neutral, fresh, etc.
   In addition, we use FHSI to develop a technique to predict the aesthetic preference for color combinations from an individual color preference and harmony. One of the possible applications of that is to to deal with the uncertainty linked to apparels images for the online shopping coordination.
   People regard color as an aesthetic issue, especially when it comes to choosing the colors for their clothing, apartment design and other objects around. Aesthetic experiences are omnipresent in modern life. However, there is no scientifically comprehensive theory that can explain, evaluate and predict aesthetic preferences. Visual stimuli are usually multidimensional, while their perception is subjective. Color is one of such dimensions and is considered to be the main variable affecting aesthetic preferences.
   Important variables involved in aesthetics are preference and harmony. FHSI space was successfully used for quantitative evaluation of the harmony and preference phenomena for the intended application of apparel coordination. Preference for color schemes is predicted by combining preferences for the basic colors and ratings of color harmony. For example, in the context of apparel coordination, it allows predicting a preference for a look based on clothing colors.
   The model was experimentally validated with three different types of experiments -Two Alternative Forced Choice, Rank Ordering and Rating. According to experiments, the model results in useful predictions of ratings of harmony and preference - the predictive power was quite high in all sessions. In addition, we analysed the system performance based on standard recall and precision metrics. We also developed a software and a library implementing the model.
   Our system differs from traditional image retrieval systems in a number of aspects, like automated item description based on color schemes and natural query language, account for a personal variation. It has potential in a wide range of color image applications and is suitable for a number of domains, including fashion, design, marketing, and art., 本文, 128p},
 school = {三重大学},
 title = {Fuzzy Model For Human Color Perception and its Application in E-commerce：Apparel Color Coordination},
 year = {2019}
}