Fixes proposed by Prof. Dr Kogler

bibliografia.bib

@@ -279,12 +279,20 @@ @Misc{ben:09
 
 @Misc{blackice:16,
  author       = {Black Ice},
- title        = {The  {HSI} color space},
+ title        = {The {HSI} color space},
  howpublished = {\url{http://www.blackice.com/images/HSIColorModel.jpg}},
  year         = {2016},
  note         = {Last access 15/10/2016}
 }
 
+@Misc{eugster:10,
+ author       = {Simon A. Eugster},
+ title        = {The {YCbCr} color model},
+ howpublished = {\url{https://en.wikipedia.org/wiki/File:YCbCr-CbCr_Scaled_Y50.png}},
+ year         = {2010},
+ note         = {Last access 26/09/2018}
+}
+
 @TechReport{ar-face-database:98,
  title       = {The {AR} face database},
  author      = {Aleix Mart\'{i}nez and Robert Benavente},
@@ -400,7 +408,7 @@ @Article{basilio:11
 }
 
 @Conference{faria:18,
- author       = {Rodrigo Augusto Dias Faria and Roberto Hirata Jr.},
+ author       = {Rodrigo Augusto Dias Faria and Hirata Jr., Roberto},
  title        = {Combined Correlation Rules to Detect Skin based on Dynamic Color Clustering},
  booktitle    = {Proceedings of the 13th International Joint Conference on Computer Vision, Imaging and Computer Graphics Theory and Applications - VISAPP},
  volume       = {5},

cap-conclusoes.tex

@@ -14,9 +14,9 @@ \section{Final considerations}
 \label{sec:final_considerations}
 The method proposed by~\citet{brancati:17} is a novel rule-based skin detection method that works in the YCbCr color space based on correlation rules that evaluate the combinations of chrominance Cb, Cr values to identify human skin pixels depending on the shape and size of dynamically generated skin color clusters (trapezoids). The method is surprisingly simple and clever and it established a new tier.
 
-Because the authors left the code available, we reproduced the original experiments and also checked if the same color patterns were presented in RGB, HSV, and Lab color spaces, or other applications as finding tree leaves, but the results were not consistent as the original approach for human skin using YCbCr space.
+Because the authors left the code available, we reproduced the original experiments and also checked if the same color patterns were presented in RGB, HSV, and Lab color spaces, or other applications as finding tree leaves, but the results were not similar as the original approach for human skin using YCbCr space.
 
-In this research project, we introduced two extensions based on a hypothesis that the original rule could be reversed and also taking into consideration that a human skin pixel does not appear isolated. We also made a third extension that combines the original rule with the reversed one in a more strengthen method in terms of precision. All these extensions are simple and do not hurt the efficiency of the original method.
+In this research project, we introduced two extensions based on a hypothesis that the original rule could be complemented and also taking into consideration that a human skin pixel does not appear isolated. We also made a third extension that combines the original rule with the complementary one in a more strengthen method in terms of precision. All these extensions are simple and do not affect the efficiency of the original method.
 
 We tested the extensions in four standard public datasets and the experiments showed that our methods improve the accuracy of skin detection, even when there exists a huge variation in ethnicity and illumination. Moreover, our approach proved to be very competitive, outperforming alternative state-of-the-art work. In addition, we implemented a grid search for the parameters tuning and the supplementary neighborhood operations as well as new tools to process the datasets, binarize images, and a visualization web application for the problem of skin detection\footnote{Available at \url{https://bitbucket.org/rodrigoadfaria/skin-detector-ws}.}.
 

cap-introducao.tex

@@ -35,7 +35,7 @@ \chapter{Introduction}
 
 It is worth mentioning that image processing is one of the most important tasks in a computer vision system. Its goal is to create a suitable description -- typically based on shapes, textures, gray levels or color -- with enough information to differentiate the objects in the scene. With this description, useful interpretation can be extracted from the image by means of an automatic computer system that facilitates human perception~\citep{gonzalez:02}.
 
-There is no general agreement among authors regarding where image processing stops and computer vision starts. The first, as the title says, processes the image by applying some transformations on it which will produce a more enhanced and readable image. In addition, the input and output of the process are always images. On the other hand, computer vision has the ultimate goal to use computers to emulate human vision, including learning and the ability to make inferences and take actions based on visual inputs. In other words, computer vision is intended to, based on images, obtain more abstract representations~\citep{gonzalez:02}. In general,
+There is no general agreement among authors regarding where image processing stops and computer vision starts. The first, as the title says, processes the image by applying some transformations on it which will produce a more enhanced and readable image. In addition, the input and output of the process are always images. On the other hand, computer vision requires the control of acquisition and decisions about the visual input process and, requires a visual agent which performs this control~\citep{gonzalez:02}. In general,
 computer vision systems benefit from image processing techniques as pre-processing steps to build better applications. Thus, we can see that they definitely are not different fields, but there is an overlapping between them.
 
 This work is intended to explore new methods on human skin detection. We will use techniques from both image processing and computer vision fields. Color space transformation from image processing, for example, as well as human skin segmentation and understanding as part of computer vision. This is a tentative to imitate the human visual system and its capability to recognize others from the same species -- of course, humans use other characteristics to identify other humans like shape, height, gender, and others, but the skin is also part of this recognition system.
@@ -44,15 +44,15 @@ \chapter{Introduction}
 
 The human skin color pixels have a restricted range of hues and are not deeply saturated since the appearance of skin is formed by a combination of hemoglobin (red) and melanin (brown, yellow), which leads the human skin color to be clustered within a small area in the color space~\citep{fleck:96}.
 
-Color has the ability of functioning as a descriptor that often simplifies the identification and extraction of an object in a scene. Moreover, the ability of humans to discern thousands of tonalities and intensities compared to only a few dozen levels of gray put the color as a strong candidate feature in computer vision and image processing applications~\citep{gonzalez:02}.
+Color has the ability of functioning as a descriptor that often simplifies the identification and extraction of an object in a scene. There are thousands of tonalities and intensities compared to only a few dozen levels of gray, which puts the color as a strong candidate feature in computer vision and image processing applications~\citep{gonzalez:02}.
 
 In general, the colors are represented by their brightness, hue, and saturation, which are usually the features used to distinguish one color from another. The brightness gives the notion of chromatic intensity. Hue represents the dominant color perceived by an observer. Saturation refers to the relative purity or amount of white light applied to the hue. Combined, hue and saturation are known as chromaticity and, therefore, a color must be characterized by its brightness and chromaticity~\citep{gonzalez:02}.
 
 Colors can be specified by mathematical models in tuples of numbers in a coordinate system and a subspace within that system where each color is represented by a single point. Such models are known as the color models~\citep{gonzalez:02}.
 
-The choice of a color space is also a key point of a feature-based method when using skin color as a detection cue. Due to sensitivity to illumination in the scene, the input image is, in general, first transformed into a color space whose luminance and chrominance components can be separate to mitigate the problem~\citep{vezhnevets:03}.
+The choice of a color space is also a key point of a feature-based method when using skin color as a detection cue. Due to sensitivity to illumination in the scene, the input image is, in general, first transformed into a color space whose luminance and chrominance components can be separated to mitigate the problem~\citep{vezhnevets:03}.
 
-For the case of skin detection methods, there are, basically, three approaches: rule-based, machine learning based, and hybrid. They differ in terms of classification accuracy and computational efficiency. Machine learning and hybrid methods require a training set, from which the decision rules are learned. In general, such approaches outperform the rule-based methods but require a large and representative training dataset as well as it takes a long training time and, regularly, a costly classification time as well, which can be a deal breaker for real-time applications~\citep{kakumanu:07}.
+For the case of skin detection methods, there are, basically, three approaches: rule-based, machine learning based, and hybrid. They differ in terms of classification accuracy and computational efficiency. Machine learning and hybrid methods require a training set, from which the partition configuration is learned. In general, such approaches outperform the rule-based methods but require a large and representative training dataset as well as it takes a long training time and, regularly, a costly classification time as well, which can be a deal breaker for real-time applications~\citep{kakumanu:07,brancati:17}.
 
 In this work, we propose an improvement of a novel method for rule-based skin detection that works in the YCbCr color space~\citep{brancati:17}. Our motivation is based on the hypothesis that the original rule can be complemented by another rule that is a reversal interpretation of the one originally proposed. Besides that, we also take into consideration that a skin pixel does not appear isolated, so we propose another variation based on neighborhood operations. The set of rules evaluate the combinations of chrominance Cb, Cr values to identify the skin pixels depending on the shape and size of dynamically generated skin color clusters \citep{brancati:17}. The method is very efficient in terms of computational effort as well as robust in very complex images (i.e. images with varied backgrounds, numerous colors, textures, and shapes).
 
@@ -63,7 +63,7 @@ \section{Motivation}
 
 The subject of the research has been something that was attractive for us from the very beginning of the program. First, with a project on race classification in partnership with the industry. Then, with the intensification in the search for related works, with the problem of skin detection.
 
-The latter led us to the brilliant work of~\citet{brancati:17}: a new rule-based skin detection method that works in the YCbCr color space. Here, more specifically, our motivation was to propose improvements based on the hypotheses that: (1) the original rule can be reversed and, (2) human skin pixels do not appear isolated, i.e. neighborhood operations are taken in consideration.
+The latter led us to the brilliant work of~\citet{brancati:17}: a new rule-based skin detection method that works in the YCbCr color space. Here, more specifically, our motivation was to propose improvements based on the hypotheses that: (1) the original rule can be complemented and, (2) human skin pixels do not appear isolated, i.e. neighborhood operations are taken in consideration.
 
 
 %% ------------------------------------------------------------------------- %%