1. INTRODUCTION:

The skin is the outermost layer of the body. It is frequently exposed to the environment, where it may come into touch with dust, microorganisms, and UV radiation. These might be the cause of any disease. Skin-related disorders are made more complicated by genetic instability^1,2. The skin is linked to a number of skin diseases, affecting a person's appearance and capacity to operate. Skin infections are caused by bacteria, fungi, parasites, or viruses³. As depicted in Figure 1, the human skin has three layers such as Dermis, Epidermis, and Hypodermis.

Figure 1: Human Skin Anatomy

The major reason of skin diseases are the most common causes of skin diseases such as fungal infection, bacteria, allergies, viruses, and other factors. The texture or color of the skin changes in general, diseases are chronic, infections etc. Skin infections must be identified early to avoid development and spread. And it damages the patient both financially and physically⁴. Because of its complexity, dermatology is one of the most difficult fields to diagnose. The widespread use of smartphones in a developing country has opened up new opportunities for low-cost early disease diagnosis. We are obtaining image processing skills for device diagnostics using smartphone or digital camera technologies. We created an application that takes a two-stage approach to solving problems. Image processing was used in the first stage to identify the problem, and machine learning was used in the second stage to provide a solution⁵. The difficulty for the patient to diagnose is that a condition that appears to be a feature of one disease in the early stages may turn out to be a sign of another in later stages. Some disorders need medical attention, yet they all have common flaws. A machine learning model is trained on the assessed properties obtained by a microscope analysis of a skin sample to tackle this challenge as a result, the dermatologist.

The key contrubution of ths study:

i) Create Database for images for Different Types of Skin Disease.

ii) Comparative Analysis with Mobile Phone Camera and Digital Sony HD Camera.

2. RELATED WORKS.

Recently, the advanced development of technology is studied about the classification of the disease combined with digital image processing has been performed. For eczema identification, the researchers have used the SVM-based supervised learning system, multi-model, and multilevel approach for analysis⁶. The SVM was used in to diagnose some circulatory infections based on the color of the fingernails⁷. A similar approach was used to diagnose infections using melisma pictures⁸. This indicates that a solution was presented for identifying BCC (Basel Cell Carcinoma). The system is capable of accurately recognizing the existence of basal-cell carcinoma using adequate thresholding values with a percentage reliability of 91.33% in the detection. The proposed method employs, the Sobel operator for segmentation, and three distinct skin diseases are chosen: seborrhoea keratosis, pyoderma, and dermatitis⁹. Two feature sets were tested; one is color and texture characteristics, and the second is 4,182 color and texture features. The results were encouraging, the average F-Measure for the 86 feature being 86.67% and the average F-measure for the 4,182 feature being 84.12%. As a classification method, the building of an SVM has really split the dataset into different classes. Using this method, three groups of skin diseases images were categorized as skin lesion segmentation, ABCD rules, and GLCM. The data were classified using KNN, Random Forest, and SVM. The classifier showed a high accuracy of 89.93 percent when the ABCD Feature Extraction was used¹⁰. The literature has made substantial progress in the detection of skin diseases. However, the suggested approach is primarily intended for the identification of a single form of skin disease, making it difficult to apply to the exact identification of many types of skin disease. We found that previous researchers, but there is relatively little study on utilizing one approach to categorize two or more disorders. We proposed an efficient technique in which a database of pre-processed images is trained and tested and classified using image segmentation, GLCM, color extraction, LBF technique, SVM, KNN, K-Means algorithms, and machine learning-based algorithm, to determine whether the skin lesion is really useful in the diagnosis of various skin diseases using two different resolutions such as camera photos and mobile phone images. Diagnosis of skin conditions from a picture is a difficult challenge. There are so many different types of skin disorders defined as acne, psoriasis, eczema, leprosy, wrath, ringworm, and vitiligo. It is presented in table 1.

Table 1: Different Types of Skin Disease

Disease

Name

Types of Disease

Images

Symptoms

Viral skin Disease

Eczema, Psoriasis,

Vitiligo, Hives, Seborrheic, Impetigo, Cold sore, Ringworm, Lupus, Measles,

Cellulitis, Rosacea.

1. Babies,

2. Plantar warts in adults,

3. Kaposi's sarcoma in HIV’s infected patients

Signs of viral infection

1. crying, excessive

2. Sleepiness,

3. complexity feeding,

4. Bulging of the soft spot

Also at top of the head.

Fungal

skin disease

Rosacea, lupus, Vitiligo,

Melis ma, Impetigo,

Pillars, Vitiligo, SSC, Ringworm blisters.

Fungal infection, it caused by

man and women, but more common in women

Skin changes, red and

Possibly cracking and peeling skin Itching.

Pigmented skin

Disease

Vitiligo, melanocytic, naive (mole), seborrheic keratosis, lentiginous skin cancer, melanoma, and, pigment, post-inflammatory pigmentation due to past

Injury.

There are different categories of pigmented skin benign, dysplastic, and melanoma, produced by the accumulation of melanocyte cells do mention of melanocytes cell.

Vitiligo is a condition that causes patches of light skin.

Allergic skin

Disease

Dermatitis, Atopic eczema, and Leprosy.

Eczema is most common in children. A number of people who has food sensitivity that can make eczema symptoms worse.

Heavy fever, Itching and spreading and red spot on all body parts, and red bumpy, scaly, itchy or

Swollen skin at the point

Of contact.

2.1 Challenges:

· The disease has many skin lesion types.

· Many diseases may have similar characteristics, which is often confusing for the dermatologist as well as to identify the disease by graphically check.

· Classifying multiple skin lesions into a correct class is challenging due to the high similarity among different lesions.

3. METHOD AND TECHNIQUES:

Figure 2.The Block Diagram of Proposed Methodology

1.1 Description of the dataset

Table 2: Local Database for Skin Disease

Sr. No.

Name of Institute/ Organisation

Database Size

Name of Disease

Resolution

Name of Devices

Govt. Hospital (GHATI), Aurangabad Sony HD Camera.

160

Acne, psoriasis, Leprosy, Eczema, Wart, Melisma, Ringworm, Ulcer, Vitiligo, Skin Cold.

6016*3384

Sony HD camera

Govt. Hospital (GHATI), Aurangabad. One plus a Mobile Camera.

154

Acne, psoriasis, Leprosy, Eczema, Wart, Melisma, Ringworm, Ulcer, Vitiligo, Skin Cold.

3000*4000, 1156*867

Smartphone

We have taken a total of 314 images from different resolution cameras such as a Digital Sony HD Camera and a Smart Mobile Phone Camera at the Department of Dermatology and Government Hospital of Medical College and Hospital of Aurangabad. We have captured patient disease images under the observation of dermatologists. To identify images from the clinical dataset, we have taken nine skin disease datasets such as Acne (43), Psoriasis (50), Leprosy (43), Eczema (41), Wart (28), Melisma (10), Ringworm (32), Vitiligo (30), Ulcer (10), and Skin Cold (15). But in our paper we used only five types of database such as Acne, Psoriasis, Eczema, Wrath, and Ulcer. They are seen very Often in India. Skin Disease images dataset, we have collected and classified database into training and testing using python toolbox library.

3.2 Image pre-processing:

Improving image quality in the pre-processing step begins with removing unnecessary noise. Hair and bubbles may appear as noise inside the picture of the photograph. It becomes required to filter the picture to remove noise, and some photos contain an undesired feature that is occasionally present in dermoscopic images: hair. Because noise affects categorization accuracy. In this paper, we used filters such as Gaussian filter, Adaptive filter, and Median filter for denoising, which includes Gaussian noise, salt and paper noise, passion noise, and speckle noise. We discuss this in previous papers in more detail .and Gaussian filter is used to reduce noisy images. As shown in table 3.

The tests were carried out on a variety of standard photographs of various resolutions. Python programming is used to carry out reproduction. An unpleasant influence of mimicked Gaussian, salt and paper noise, speckle noise, and Poisson noise contaminates the information. As the PSNR lowers, the MSE rises, and vice versa. When the peak signal-to-noise ratio rises, the resulting image becomes highly smooth to the eyes perception, and the image returns to its previous state. Images that are highly deformed have a high value. As Described in Table 4, numerous types of noise can be found in the table.

Table 3: Performance of Skin Disease Images on Different Noise

Disease Name	MSE	PSNR	Entropy	Noise
Acne	53.67	22.651	7.75	Gaussian Noise
Psoriasis	41.20	38.25	8.26
Ulcer	31.25	36.58	6.89
Eczema	51.26	28.96	7.63
Acne	59.76	20.30	7.05	Salt and Paper Noise
Psoriasis	56.26	51.02	7.49
Ulcer	32.33	36.92	6.96
Eczema	49.23	27.70	6.36
Acne	58.73	27.49	7.43	Poisson Noise
Psoriasis	63.21	59.69	7.52
Ulcer	36.44	38.24	7.26
Eczema	33.36	29.63	7.25
Acne	56.76	20.92	5.48	Speckle Noise
Psoriasis	52.02	49.51	6.89
Ulcer	32.69	36.41	7.37
Eczema	33.22	28.63	7.44

3.3 Filtering Techniques:

We have applied Gaussian filter, Median filter, and Wiener filter to de-noised images on the noisy image, consequently obtaining the denoised image and calculating the MSE, PSNR, and Entropy values.

Figure 3: Noise Detection Historiographical

Table 4: Filter applied on Noise and Calculated PSN, MSE and Entropy values

Various Filter	MSE	PSNR	Entropy	Name of Noise
Gaussian Filter	18.67	12.81	5.26	Gaussian Noise
	30.81	27.49	5.59	Salt and pepper Noise
	27.84	26.42	6.51	Speckle Noise
Median Filter	20.06	23.64	6.37	Gaussian Noise
	30.32	27.85	6.69	Salt and pepper Noise
	23.98	16.11	7.27	Speckle Noise
`Wiener Filter	18.92	15.57	7.44	Gaussian Noise
	31.56	29.10	5.60	Salt and pepper Noise
	18.90	15.58	6.87	Speckle Noise

The MSE, PSNR, and Entropy of each examined filter, namely Gaussian, Median, and Weiner filters, are shown in Table 4. Gaussian, Speckle, Salt, and paper noises were all removed using each filter. On the Gaussian noise, the Gaussian filter performs better than other filters, with 18.67 MSE, 12.81 PSNR, and 5.26 entropy values. On salt and pepper noise, the Wiener filter is given high values, such as 31.56 MSE and 29.10 PSNR.

Table 5: Comparisons of original and filter MSE, PSNR and Entropy

Noise	Original MSE	Filter MSE	Original PSNR	Filter PSNR	Original Entropy	Filter Entropy
Gaussian Noise	44.35	23.13	31.61	20.08	6.80	5.07
Salt and Paper Noise	49.40	28.45	33.99	24.53	6.97	5.11
Poisson Noise	47.94	27.44	38.87	25.69	7.63	5.13
Speckle Noise	43.67	26.32	33.76	23.24	7.37	5.51

3.4 Gaussian Filter:

The Gaussian filter was used to smooth the image and eliminate noise from the artifact. For the influence detected and brought about by an irrelevant backdrop of pictures, a noise through a Gaussian filter is required. It is a common method for removing salt and paper noise from photos while preserving edges and being helpful in their creation [11]. Gaussian filter is a smoothed pixel according to the power-to power coefficients. The smoothing function can be expressed in equation

(1)P_Smooth ……………..(1)

Where PSmooth (x, y) and P_row (X+i, y+j) denote the raw pixel, respectively. Represents approximation Gaussian coefficient (C normal and normal) representing the normalized coefficient. Gaussian noise is found in most of the skin disease images. We have used Gaussian filter for the removal of the noise, and we got better accuracy in MSE,PSNR.Even if the skin disease image clearly displays the better-enhanced image, the PSNR values do not interpret comparable findings, and it is easy to help evaluate them. Table 5 tabulates the average PSNR and MSE values for each tested filter, as well as the computed MSE, PSNR, and Entropy. Each filter is used to eliminate the following noise types: Gaussian, Salt and pepper, and speckle. When comparing all three filters, the Gaussian and Median filters perform better for speckle noise than salt and pepper noise. Furthermore, the Median filter outperforms other filters in terms of PSNR and MSE, but only for salt and pepper noise density levels less than 30%. Among the others, the Gaussian noise and Gaussian filter provide great accuracy as shown in table 6. MSE is given high values in every skin disease and PSNR is given less than MSE values.

3.5 Image segmentation:

Image segmentation is an important component of image recognition. The goal of image segmentation is to separate the image into various segments, determining which regions require more attention than the background [12]. To segment the image of skin Disease, the K-Means Clustering approach is utilized. This method divides the data into various cluster areas depending on the closest distance between the data and the centroid of each cluster. The image segmentation result is a picture in which the foreground is boundary detection.

Figure 4: Image Segmentation Process

The post-pre-processing stage is performed on the image segmentation with K-Means Clustering findings in Fig. 3. Since these results are regarded as less than ideal and contain some noise/small objects. The approach employs binary image processes such as the Gaussian filter, noise reduction, border cleaning, the masking process, and cropping photos of skin conditions.

3.6 Contrast:

Where I and referee to the grey level difference between adjacent pixel G(i,j) is the distribution probability mainly used in described in the degree off depth computational mathematical methods Highest contrast value goes, the deeper groove vice versa.

3.7 Entropy:

Where A2 reference to the entropy, which means the quality of information which the image can change with the different textures. As A2 increase the texture of the speck would be arrange sparsely and vice versa shown in table5.

3.8 Haar Feature:

The haar features used in voila and joneses exhibit a rectangle structure it consists of four subs rectangular some examples can be seen below. The integral image f is denoted as equation 5 and fig 4. The integral F value at position [x, y], is defined as the sum of the equation of I considering all pixels located inside the rectangular are ranging [0, 0] up to and including[x, y] ¹⁷.

These images are divided into training and testing sets, where 90 images from each group are used to train the system and the remaining 10 images in each group serve as the testing set. We are used Bcc Images and apply haar feature.

Figure 5. Haar feature using histogram

While presenting results, we tried to present images having different common problems of dermoscopic images. Figure 4 shows an image of dysphasic nevi having represents disease affected region lesion with boundary represent on disease region.

3.9 Support Vector Machine (SVM):

The SVM is a machine learning method that uses statistical theory to learn ¹⁸. When compared to other machine learning algorithms in the literature, SVM performs better than others. SVM handles limited quadratic differentiation between two classes, and it can also solve multiclass problems. The SVM method optimizes the distance between data points and hyper planes. The Support Vector Machine figure 5. Is shown below.

Figure 6. Support vector machine using Linear Regression.

A linear kernel is also one of the simplest of all the kernels available. When we want to classify two classes, which are having more features in common, then other kernels such as Polynomial Kernels are to be used to achieve better accuracy and precision ¹⁹. Gamma kernel is used to define the boundary values of the SVM ²⁰. Linear kernel’s equation is as follows:

This equation involves calculating the inner products of a new input vector (x) with all support vectors in training data ²¹. The coefficients ‘y’ is the distance from the hyper plane to the feature and ‘c’ is an optional constant ²². All the values, which were added into a database as mentioned, are given as input to the SVM and is trained to differentiate the classes as it is labeled data.

3.10 K-Nearest Neighbour (KNN):

K-Nearest Neighbour is one of the simplest Machine Learning algorithms based on the Supervised Learning technique. K-NN algorithm stores all the available data and classifies a new data point based on the similarity. This means when new data appears then it can be easily classified into a good suite category by using K- NN algorithm²³.

3.11 Feature Extraction:

There are several unique features that distinguish four classes of human skin disease, such as the texture feature and color features. These features were selected for classification.

3.12 Color Moments:

Skin disease images also can be divided based on the color variation of each class. The color feature extraction method proposed is the color moments. Color represents a solid representation of color features in characterizing image color ¹³. Color moments assume the color distribution of an image as a probability distribution. This study will use two moments from the color probability distribution. Of the image, such as mean and standard deviation. The mean standard deviation can be calculated by using the following equation ¹⁴.

In this research ¹⁵, the color method that was tested had the best accuracy to recognize features of skin disease. The color space of YCbCr is a color space component (Y, Cb, and Cr) is a color that is applied in the photography system ¹⁶, while Cb and C represent it represents red and blue.

Table 6: Color Feature with Different Classifier

Disease Name	HD Sony Camera				Mobile Phone Camera
Disease Name	Texture Feature	SVM	Hopkin	Elbow	Texture Feature	SVM	Hopkin	Elbow
Acne	79	91	82	78	62	93	86	6
Eczema	56	89	67	72	63	100	62	62
Wart	31	85	46	46	25	100	25	75
Psoriasis	50	87	70	40	66	89	88	67
Ulcer	57	90	48	52	76	95	67	76

Figure 7: Comparison of Sony HD Camera and Mobile Camera dataset

We have used in color feature, SVM, Hopkin and Elbow techniques for classification of skin disease such as Eczema, Acne, Wrath, Psoriasis, Ulcer. It has given different accuracy in different skin disease. We are shown in table 6. SVM is given good accuracy in the Acne (94%), wart (83%), eczema (92%) and ulcer (95%) disease and color feature is given 100% accuracy in Psoriasis disease. The mobile dataset is given best accuracy in SVM techniques. Hopkin and elbow techniques are given less result for skin disease.

3.13 Texture Feature Extraction:

Compare with the traditional way, GLCM effective tool analysis features of texture, the textures of different type of disease can be obtained entropy, contrast, in this paper, there are used in nine different types of disease are selected as main research object which is Acne, Psoriasis, Leprosy, Eczema, Warth, Melisma, Ringworm, Vitiligo, and Ulcer, Skin Cold, respectively. All pictures as shown in Figure 5. Are extracted from photos of a government hospital.

Table 7: Texture Feature with Different Classifier

Disease Name	HD Sony Camera				Mobile Phone Camera
Disease Name	Texture Feature	SVM	Hopkin	Elbow	Texture Feature	SVM	Hopkin	Elbow
Acne	79	91	82	78	62	93	86	6
Eczema	56	89	67	72	63	100	62	62
Wart	31	85	46	46	25	100	25	75
Psoriasis	50	87	70	40	66	89	88	67
Ulcer	57	90	48	52	76	95	67	76

Figure 8: Result in Different Techniques

Figure 9: Confusion matrix of SVM

We have used Texture Feature and techniques for skin disease such as Eczema, Acne, Wrath, Psoriasis, and Ulcer. It has given different accuracy in different skin disease, it shown in table 7. The SVM is given best accuracy for Eczema 100%, Wart 100%, Ulcer 95%, Acne 93%, and Psoriasis 89% in mobile dataset.

4. PERFORMANCE MEASURES:

A confusion matrix is used for computing performance measures of classification for skin disease such as accuracy, sensitivity, specificity, positive predictive value, and negative predictive value a, a confusion matrix contains is given information about actual and predicted classifications. We have also calculated precision, recall, and F1 score for each disease (Table 8.), to guard against bias caused by the unbalanced distribution of different disease.

Table 8: Summary of Precision-Recall and F1-Score for skin Disease classification.

Disease No.	Precision	Recall	F1score
Acne	1.00	1.00	1.00
Acne Nod	0.67	0.67	0.67
Psoriasis	0.8	0.8	0.8
Eczema	1.00	1.00	1.00
Warth	0.89	0.89	0.89
Ulcer	0.6	0.6	0.6

We have observed the result and we got 1 precision accuracy for Acne and, Eczema disease and less accuracy for Ulcer (0.6).

5. RESULT AND DISCUSSION:

The train and test ratio is important factor affecting the classification result. It is observed that the training set size increases, the results is improved. The effect of train / test classification accuracy is studied and the best classification results are reached with 80/20 % train and test ratio. We are observed that, overtraining may also lead to less accuracy. We not only performed classification but also took a step forward and tried to classify all 314 unique sub-classes as well. We found 100% accuracy for Color Feature and Haar feature in training data and 99.4 and 97.23 % accuracy in test dataset. We have got less accuracy for texture feature in training and testing dataset 86.23 % and 75% respectively.

Table 9: Feature Type of Skin Disease

Feature Type	Accuracy
Feature Type	Train data	Test data
Color feature	100%	99.4%
Texture feature	86.23%	75%
Haar feature	100%	97.23%

The colour moment approach was used to extract colour features from a variety of colour spaces. The colour space of the picture to be examined is RGB, HSV, and YCbCr. The three colour spaces are tested for correctness. The RGB, HSV, and YCbCr colour feature types exhibit the accuracy gained after the feature extraction experiment using the Color moments technique in the colour space. Color space excludes the influences of light, which is alter the characteristics of skin colour, allowing for the extraction of a wealth of feature information. Based on test results, some haar varieties are more accurate than others.

Table 10: Classification Result of Mobile Camera and Sony HD Camera

	Mobile Camera				Sony HD Camera
Features	SVM	K-NN	K-Means		SVM	K-NN	K-Means
Features	SVM	K-NN	Elbow	Hopkins	SVM	K-NN	Elbow	Hopkins
Haar Feature	92%	89%	86%	88%	91%	85%	86%	88%
FCM	92%	90%	86%	86%	89%	78%	86%	86%
OS-FCM	92%	87%	88%	88%	91%	86%	88%	88%
GLCM	90%	87%	86%	87%	88%	81%	86%	87%
LBF	89%	88%	86%	87%	87%	78%	86%	86%
Color Feature	89%	87%	88%	87%	89%	87%	88%	88%

The system is produced high accuracy when applied multiple skin disease classes with the help of mobile camera and Sony HD camera Camera resources. The result show that the proposed system correctly identified patient disease. The system is correctly identified Acne, Eczema, posryasis, Warth, and Ulcer skin disease. We are used different types of classifiers SVM, KNN, and K-Means and used Haar feature, color feature, FCM, OS-FCM, GLCM and LBF features for classifications. The classification algorithm is developed to predict diagnosis system. The Haar feature, FCM, and OS-FCM are given good result in SVM techniques for mobile camera dataset. GLCM feature is given 90% accuracy in SVM. LBF and color feature are given 89% accuracy in SVM technique. Sony HD camera dataset is also given good result in SVM technique for used Haar feature, color feature, FCM, OS-FCM, GLCM and LBF features as shown in table 10. The proposed model is given 92 % accuracy. The SVM classifier is also maintained a substance score.

6. CONCLUSION:

The early diagnosis of skin disease system by using computer aided techniques. This method is combined artificial intelligent and digital image processing for skin disease and identify the skin disease. We have collected 314 dataset from Government Hospital, Aurangabad. We have taken five skin disease such as Acne, Psoriasis, Eczema, Warth, and Ulcer. We have used Gaussian noises and Gaussian filter pre-processing techniques for remove the image noise and enhance image quality. We are used different types of classifiers SVM, KNN, and K-Means and used Haar feature, color feature, FCM, OS-FCM, GLCM and LBF features for classifications. SVM is given better accuracy for mobile camera dataset. SVM is proved more succeeded in doing image recognition. It gives the 92% accuracy. The KNN and K-Mean are given less accuracy.

7. REFERENCES:

1. R. J. Hay, N. E. Johns, H. C. Williams, I.W. Bolliger, R. P. Dellavalle, and D. J. Margolis, “The global burden of Skin disease in 2010: An Analysis of the prevalence and impact of skin conditions”, 55 J, Investigative Dermatology, vol. 134, no. 6, PP NO.1527_1534, 2014.

2. Sardana K, Mahajan S, Sarkar R. Spectrum of skin diseases among Indianchildren. Pediatr Dermatol, 26 (1) pp no:6–13,2009.

3. Palak Mehta, Bhumika Shah, “Review on Techniques and Steps of Computer Aided Skin Cancer Diagnosis” International Conference on Computational Modeling and Security (CMS2016). https://doi.org/10.1016/j.procs.2016.05.28.

4. Online available: https://www.medicalnewstoday.com/articles/154322.

5. Abraham Getachew Kelbore, Philip Owiti, Anthony J. Reid, Efa Ambaw Bogino, Lantesil Wondewo sen and Blen Kassahun Dessu, “Pattern of skin disease in childeren attending a dermatology clinic in a referral hospital in wolaita soda, southern Ethiopia”, BMC Dermatology, https://doi.org/10.1186/s12895-019-0085-5, pp no- 3-8, 2019.

6. Housman TS, Feldman SR, Willi ford PM, Fleischer AB Jr., Goldman ND, et al., “Skin cancer is among the Most Costly of all Cancers to treat for the Medicare population”, J Am Acad Dermatol 48: pp. 425_429, 2003.

7. Arifin, S., Kibria, G., Firoze, A., Amini, A., and Yan, H. (2012) “Dermatological Disease Diagnosis Using Color-Skin Images.” Xian: International Conference on Machine Learning and Cybernetics.

8. Yasir, R., Rahman, A., and Ahmed, N. (2014) “Dermatological Disease Detection using Image Processing and Artificial Neural Network.“Dhaka: International Conference on Electrical and Computer Engineering.

9. Santy, A., and Joseph, R. (2015) “Segmentation Methods for Computer-Aided Melanoma Detection.”Global Conference on Communication Technologies.

10. Zeljkovic, V., Druzgalski, C., Bojic-Minic, S., Tameze, C., and Mayorga, P. (2015) “Supplemental Melanoma Diagnosis for Darker Skin Complexion Gradients.” Pan American Health Care Exchanges

11. Suganya R. (2016) “An Automated Computer-Aided Diagnosis of Skin Lesions Detection and Classification for Dermoscopy Images.”

12. Alam, N., Munia, T., Tavakolian, K., Vasefi, V., MacKinnon, N., and Fazel-Rezai, R. (2016) “Automatic Detection and Severity

13. Kumar, V., Kumar, S., and Saboo, V. (2016) “Dermatological Disease Detection Using Image Processing and Machine Learning.” IEEE.

14. A.A.L.C. Amarathunga, E.P.W.C. Ellawala, G.N. Abeysekara, C. R. J. Amalraj, “Expert System for Diagnosis of Skin Diseases”, International Journal of Scientific and Technology Research, IEEE, PP No-1-5, VOLUME 4, ISSUE 01, ISSN 2277-8616, January 2015.

15. Krizhevsky, A., ILYA, S., and Geoffrey, E. (2012) “ImageNet Classification with Deep Convolutional Neural Networks.” Advances in Neural Information Processing Systems.

16. Cristianini, N., Shawe, J., “Support Vector Machines”, 2000.

17. SOMMERVILLE, I., “Software Engineering”. 9th .2011.

18. Smits, Guido F., and Elizabeth M. Jordaan. "Improved SVM regression using mixtures of kernels." Proceedings of the 2002 International Joint Conference on Neural Networks. IJCNN'02 (Cat. No. 02CH37290). Vol. 3. IEEE, 2002.

19. Cesar Souza (2010, March 17) “http://crsouza.com/2010/03/17/kernelfunctions-for-machine-learning-applications.

20. Fatima, Ruksar, Mohammed Zafar Ali Khan, and K. P. Dhruve. "Computer-aided multi-parameter extraction system to aid early detection of skin cancer melanoma." International Journal of Computer Science and Network Security 12.10 (2012): 74-86.

Received on 27.04.2022 Modified on 10.05.2022

Research J. Science and Tech. 2022; 14(3):145-155.

DOI: 10.52711/2349-2988.2022.00024