Modified Majority Voting Algorithm towards Creating Reference Image for Binarization

Performance Evaluation of Multiple Image Binarization Algorithms Using Multiple Metrics on Standard Image Databases

“…The choice of certain size portioning method [1] sub divides each image if sh d not support for improving our reference image creation e of experiments, two experiments are proven to be giv nstructing ground truth image. The first measure is 3 cribed above [12] in which total seven binarization meth hod also not free from failure of creating proper refere RI and texture. nce image corresponding is different type of original image nce image by 30% deviation, c) Arial image and d) its equiva ation, e) Texture image and f) its equivalent reference image in image, h) its equivalent reference image by 30% deviation esent experiment, the reference image has been created s described in [1] but with total 7 binarization algorit w methods i.e.…”

“…A measure is proposed in [12], in which total seven binarization method has been used i.e. Otsu, Niblack, Bernsen, Sauvola, Th-Mean, Kapur and iterative as a framework, the threshold of each algorithm is being calculated and only those algorithms are take to majority voting for reference image creation in which its threshold doesn't deviate more than deviational parameter from the average value of all threshold values pre-calculated but this method also not free from failure of forming proper reference images mainly for Arial, MRI and texture.…”

Integration of Decision Trees Using Distance to Centroid and to Decision Boundary

“…This level represents the information that the classifier Ψ assigns the unique class label ω c to a given recognized object x, i.e. the output of the base classifier indicates uniquely the class label [Dey et al, 2014], [Przyby la-Kasperek and Wakulicz-Deja, 2017]. The other most commonly used type of the classifier output is the score function that addresses the degree of assigning the class label to the given recognized object x.…”

“…These decision trees are the base classifiers for the considered case of EoC. If all K base classifiers are equal contribution to make the final decision of EoC and the abstract level is considered, then the majority vote rule can be applied [Fechner and Keller, 2004], [Mohandes et al, 2018]. This method allows counting base classifiers outputs as a vote for a class and assigns the input pattern to the class with the greatest count of votes.…”

“…Ensemble methods (an ensemble of classifiers EoC) [Giacinto and Roli, 2001], [Ponti Jr, 2011], [Przyby la-Kasperek and Wakulicz-Deja, 2017] are a popular approach in building a classification model that is more stable and a model that uses a set of many individual classifiers (base learners) and combine them to classify new data [Le et al, 2013], [Rokach, 2010]. The main concept behind the ensemble technique is to create a classification method that outperforms every one of the base classifiers and, as it was previously mentioned, the outputs of the base classifiers can be used in a variety of ways to determine the decision of a classifiers committee.…”