JURNAL RISET INFORMATIKA 
Vol. 5, No. 1. December 2022 

P-ISSN: 2656-1743 |E-ISSN: 2656-1735 
DOI: https://doi.org/10.34288/jri.v5i1.488 

Accredited rank 3 (SINTA 3), excerpts from the decision of the Minister of RISTEK-BRIN No. 200/M/KPT/2020 

 

 
105 

 

 

TEXTURE FEATURE EXTRACTION OF PATHOGEN MICROSCOPIC IMAGE 
USING  DISCRETE  WAVELET TRANSFORM 

 
Hasan Basri 

 
Information System 

Institut Teknologi dan Bisnis Bank Rakyat Indonesia 
Jakarta, Indonesia 

http://bri-institute.ac.id/ 
basrihasan@bri-institute.ac.id 

 
 

Abstract 
This study used a case study of Jabon leaves, and the pathogen is one of the causes of disease that attack the 
leaves of jabon, one of the leaf spots and leaf blight. Discovery of leaf spot disease in different pathogens and 
leaf blight. The pathogen was obtained from the leaf spot of Curvularia sp. 1 and Pestalotia sp., while the 
pathogen came from Curvularia sp. 2 and Botrytis sp. Identify the pathogen as soon as possible to minimize 
its effects. Improper handling can lead to increased virulence and resistance to the pathogen. Improper 
handling also can cause a disease outbreak (disease epidemic) in a region. This study is the first step in 
identifying the pathogens responsible for Jabon leaf disease. In this study, the Application of Koch's 
Postulates method to achieve the purification of pathogens and retrieve the microscopic pathogen image as 
the data acquisition stage. Furthermore, use of the segmentation stage to separate the object pathogen from 
the background, and one of the methods used is Otsu Thresholding. The extraction process of pathogen 
microscopic image using Discrete Wavelet Transform (DWT), DWT extraction results can be obtained using 
energy and entropy information. 
 
Keywords: DWT; pathogen; extraction feature 
 

Abstrak 
Penelitian ini menggunakan studi kasus daun jabon, Patogen merupakan salah satu penyebab penyakit yang 
menyerang daun jabon, salah satunya bercak daun dan hawar daun. Penemuan penyakit bercak daun pada 
patogen yang berbeda begitu pula pada penyakit hawar daun. Patogen diperoleh dari bercak daun Culvularia 
sp. 1 dan Pestalotia sp., sedangkan patogen yang berasal dari hawar daun Culvularia sp. 2 dan Botrytis sp. 
Patogen harus diidentifikasi sesegera mungkin untuk meminimalkan efek yang ditimbulkannya. Penanganan 
yang tidak tepat dapat menyebabkan peningkatan virulensi dan resistensi patogen. Penanganan yang tidak 
tepat juga berpotensi menimbulkan wabah penyakit (wabah penyakit) di suatu wilayah. Tujuan dari 
penelitian ini adalah tahap awal untuk mengidentifikasi patogen yang ada pada penyakit daun jabon. Pada 
penelitian ini, metode Postulat Koch diterapkan untuk mendapatkan purifikasi patogen dan kemudian 
mengambil citra patogen mikroskopis sebagai tahap akuisisi data. Selanjutnya adalah tahap segmentasi yang 
digunakan untuk memisahkan objek patogen dengan backgroundnya, sedangkan salah satu metode yang 
digunakan adalah Otsu Thresholding. Proses ekstraksi citra mikroskopis patogen menggunakan Discrete 
Wavelet Transform (DWT), hasil ekstraksi DWT dapat diperoleh dengan menggunakan informasi energi dan 
entropi. 
 
Kata kunci: DWT; pathogen; fitur ekstraksi 
 
 

INTRODUCTION 
 
Jabon Merah (Neolamarckia Macrophylla 

(Wall.) as a timber plant has the potential to meet 
the national timber demand (Larekeng, Qalbi, 
Rachmat, Iswanto, & Restu, 2022). Several diseases 
attack the jabon, including root, stem, and leaf 
disease. Most jabon disease is caused by fungi 

(Warisno & Dahana K, 2011). The fungus that 
attacks the leaves of jabon causes spots, blight, and 
dead buds. The fungus has several properties, one of 
which is that it causes plant diseases. These fungi 
are categorized as pathogens (Agrios G, 2005). The 
spread of pathogens is very broad (Hadi S, 2001). 
Jabon seeds are among the most vulnerable to 
pathogen attacks (Aisah, 2014). It is necessary to 



P-ISSN: 2656-1743 | E-ISSN: 2656-1735 
DOI: https://doi.org/10.34288/jri.v5i1.488 

JURNAL RISET INFORMATIKA 
Vol. 5, No. 1. December 2022 

Accredited rank 3 (SINTA 3), excerpts from the decision of the Minister of RISTEK-BRIN No. 200/M/KPT/2020 

 

 
106 

 

 

handle this as early as possible, so the growth of 
jabon is not hampered. 

According to Herliyana, the fungi have an 
abundance of type, only known about 98.998 
species, and which are a pathogen number of about 
8000 species (Herliyana, 2013). Herliyana et al. 
obtained several types of pathogens from red jabon 
leaves, a pathogen found in spotting diseases, 
namely Pestalotia sp. and Rhizoctonia sp., in blight 
diseases, namely Lasiodiplodia Theobromae, 
Fusarium sp., Colletotrichum sp., Marsonia sp., 
Gleosporium sp. 1 and Gleosporium sp. 2 
(Herliyana, Sakbani, Herdiyeni, & Munif, 2020). 
Disease control should be adjusted to consider the 
type of pathogen. One way to know the pathogen is 
by conducting the pathogen identification process, 
which is done by using the aid of microscope 
magnification (Streets, 1972). 

Sudarsan et al. researched Microscope 
Image Processing (MIP) and developed a CWT-
based computer vision algorithm to characterize 
soil particle size from digital images captured with 
a microscope. The wavelet technique's efficacy in 
detecting an image's particle size is promising, and 
the portability of the image acquisition device 
produces excellent proximate soil sensors 
(Sudarsan, Ji, Adamchuk, & Biswas, 2018). Santosh 
et al. also research MIP. The discrete wavelet 
transform (DWT) is comparable to a microscope 
because distinct signal components can be 
distinguished by simply adjusting the focus 
(Santosh & Barpanda, 2020).  

This study, using primary data, collected a 
2043 microscopic pathogen image of jabon leaf 
disease that is spotting and blight, then extracted 
using the Discrete Wavelet Transform (DWT) 
method.  

 
Table 1. Morphological and texture features of the 

pathogen 
Pathogen Characteristic 

 

Culvularia sp 1. is a dark-
colored Conidia with clear tip 
cells. 3-to-5-cell Conidia, 
characterized by curving and 
enlarged central cells. 
Culvularia sp one is derived 
from jabon leaf spot disease. 

 

Culvularia sp.2. The difference 
between Culvularia sp. one and 
Culvularia sp.2. In Culvularia 
sp.1, each cell constraint has a 
light black color, while 
Culvularia sp.2 has two cell 
boundaries in the center of 
wider black Conidia. The origin 

Pathogen Characteristic 
of the disease is the pathogen of 
jabon leaf blight. 

 

Pestalotia sp. has dark Conidia 
with some clear-colored cells. 
The tip of the cell is tapered, 
with two or more tails. 
Pestalotia sp. can cause leaf spot 
disease jabon. 

 

Botrytis sp. has long-shaped 
Mycelium features, cylindrical, 
clear color, and irregular 
branches. Conidia surround 
Conidiaphore. Color Conidia 
clear or grey, cell 1. Botrytis sp. 
is one of the causes of leaf blight 
Jabon 

 
The number of pathogens is very high; 

therefore, to determine the type of pathogen, a 
researcher must manually examine the manual to 
determine the pathogen's characteristics and 
identify similar characteristics. Each image of a 
pathogen using computer vision can be identified 
because pathogens have different characteristics s, 
as shown in Table 1. 

Bangun et al. conducted another study to 
identify the microscopic pathogen image of the 
white jabon leaf. That is pathogen Colletotrichum 
sp., Curvularia sp., and Fusarium sp. using 
morphological characteristic extraction methods 
and producing the best features of compactness and 
roundness for classification (Bangun, Herdiyeni, & 
Herliyana, 2016). 

 

 
Figure 1. Cutting only one pathogen 
 

In the study above by Bangun et al., the image's 
distinguishing features were obtained by cutting 
only the pathogenic image, as shown in Figure 1. 
However, the entire image cannot be extracted. In 
this study, we attempted to extract colony image 
pathogens from a full image without cutting them 
individually.  



JURNAL RISET INFORMATIKA 
Vol. 5, No. 1. December 2022 

P-ISSN: 2656-1743 |E-ISSN: 2656-1735 
DOI: https://doi.org/10.34288/jri.v5i1.488 

Accredited rank 3 (SINTA 3), excerpts from the decision of the Minister of RISTEK-BRIN No. 200/M/KPT/2020 

 

 
107 

 

 

Other challenges of the microscopic image 
are complex images, objects that can be stacked, 
blurred images in part because the objects are 
stacked, and many other objects that become noise. 
 

RESEARCH METHODS 
 
The method proposed in this research 

includes three stages: data acquisition, 
preprocessing, and feature extraction. 

 
Data Acquisition 

This process was done in the forest 
pathology laboratory of the silviculture department 
of the IPB forestry faculty. The sample of jabon leaf 
studied is a jabon leaf from "persemain permanent 
IPB." The acquisition phase of this pathogen 
microscopic image data uses the Postulat Koch 
method to obtain pure pathogens (Streets, 1972) 
and then photographed using the Optilab camera 
with 400x. Magnification and a maximum resolution 
of 5 megapixels for the image produced under 1 
megabyte. 

 

 
Figure 2. Stage of image capture and sick leaf 

selection 
 

 
Figure 3. purification of pathogen isolation and 

pathogen image capture 
 

Figures 2 and 3 depict the various stages of 
Postulate Koch's method for identifying foliar 
disease-causing pathogens. 
 
Preprocessing 

The preprocessing step is to select the 
image to be extracted, some images with good 
pathogen image quality are selected, and the next 
process is to crop the selected image of good quality 
to remove the overly dominant noise. 

 

 
Figure 3. Cropping Culvularia sp. 1 to remove the 

dominant noise 
 

During Postulate Koch's, dominant noise 
occurs due to shooting and several other factors. So 
as in Figure 4, noise that is too dominant must be 
removed because it will greatly affect the value of 
the feature extraction of pathogen image features. 

The image is segmented to get the object of 
a microscopic pathogen image separated from its 
background, and the separator of the pathogen 
microscopic image with its background is one of 
them using the Otsu thresholding method, Otsu 
thresholding is the optimum method of global 
Thresholding (Rafael C. Gonzalez & Woods, 2008). 
This method maximizes the differences between the 
two regions using discriminant analysis (Otsu & N., 
1996) (Naga Kiran D & Kanchana V, 2019). In this 
part, Otsu Thresholding will generate a mask to get 
the microscopic object of the pathogen with 
black/white background, as in Figure 5. 

 

 
Figure 5. Segmentation stages, create a mask using 

Otsu Thresholding 
 

 
Figure 6. result of pathogen segmentation 

 



P-ISSN: 2656-1743 | E-ISSN: 2656-1735 
DOI: https://doi.org/10.34288/jri.v5i1.488 

JURNAL RISET INFORMATIKA 
Vol. 5, No. 1. December 2022 

Accredited rank 3 (SINTA 3), excerpts from the decision of the Minister of RISTEK-BRIN No. 200/M/KPT/2020 

 

 
108 

 

 

The method of Otsu Thresholding is a rapid 
method of computation. It very well separates 
microscopic objects of pathogens against the 
background with the condition that the intensity of 
the pixel of a pathogen microscope object with its 
background has a pixel intensity value that is not too 
close. Figure 6 depicts the results of Otsu 
segmentation, which produces a pathogen object 
image with greater prominence for the subsequent 
extraction procedure.  

 
Feature Extraction DWT 

DWT is used to extract image texture 
characteristics using the Scaling function and 
Wavelet function (Madhu & Kumar, 2022). An image 
can be extracted to produce four coefficients. 
 

𝑊𝜑 (𝑗0, 𝑚, 𝑛)   = 
1

√𝑀𝑁
 ∑ ∑ 𝑓(𝑥, 𝑦)𝑁−1𝑦=0

𝑀−1
𝑥=0 ∗ 𝜑𝑗0,𝑚,𝑛 (𝑥, 𝑦)  ...  

 ......................................................................................... (1) 

𝑊𝜓
𝑖 (𝑗, 𝑚, 𝑛)   = 

1

√𝑀𝑁
 ∑ ∑ 𝑓(𝑥, 𝑦)𝑁−1𝑦=0

𝑀−1
𝑥=0 ∗

𝜓𝑖
𝑗,𝑚,𝑛

(𝑥, 𝑦),   𝑖 = {𝐻, 𝑉, 𝐷}  .................................. (2) 

 
Four coefficients from DWT extraction can 

be analyzed: LL, LH, HL, and HH. DWT can analyze 
the information of an object in the image with the 
method of multiresolution analysis (MRA), which 
was introduced by Stephane Mallat (Rafael C. 
Gonzalez & Woods, 2008). MRA is used to analyze 
the depth of information contained in an image. At 
this stage, the image is processed using DWT with 
levels 1 – 4 of decomposition (multiresolution). 

Some extraction images using DWT can be 
seen from the different spectrums of each image, 
showing that each pathogen has different texture 
characteristics when extracted using DWT. The LL 
coefficient shows the approximation of the original 
image. As for the other results of the LL coefficient, 
it can reduce or eliminate small noises in the image 
so that the LL coefficient gives the value of a cleaner 
image that is close to the pure value of an image. The 
higher the DWT decomposition level of an image, 
the LL coefficient is reduced in size and 
approximates the image approximation. 

The coefficients of LH, HL, and HH can each 
show the texture value of an image and the edge of 
an image based on horizontal, vertical, and diagonal 
orientation. LH, HL, and HH coefficients at DWT 
levels increasingly show an edge and image 
characteristic. As for DWT extraction can be used 
some family wavelet that is Daubechies, Symlet, and 
Coiflet. Wavelet family is different only on the value 
of high filter and low filter just in processing an 
image but can show different results in any retrieval 
of information contained in the image. Figures 7, 8, 
9, and 10 display examples of the spectrum images 

of each pathogen studied. We can see that each 
studied pathogen has a unique spectrum. 

 

 
Figure 7. DWT extraction results from the 

pathogen coefficients LL, LH, HL, and HH in the 
Culvularia sp. 1 spectrum  

 
 

 
Figure 8. DWT extraction results from the 

pathogen coefficients LL, LH, HL, and HH in the 
Botrytis sp. Spectrum 

 

 
Figure 9. DWT extraction results from a pathogen, 

coefficients LL, LH, HL, and HH in the Culvularia sp. 
2 spectrum 

 

 
Figure 10. DWT extraction results from the 

pathogen coefficients LL, LH, HL, and HH in the 
Pestalotia sp. spectrum 



JURNAL RISET INFORMATIKA 
Vol. 5, No. 1. December 2022 

P-ISSN: 2656-1743 |E-ISSN: 2656-1735 
DOI: https://doi.org/10.34288/jri.v5i1.488 

Accredited rank 3 (SINTA 3), excerpts from the decision of the Minister of RISTEK-BRIN No. 200/M/KPT/2020 

 

 
109 

 

 

Image extraction results using DWT then 
calculated the information contained therein using 
entropy and energy. Energy is the sum of the 
absolute value of data. The energy can present the 
texture image characteristic, and the energy can be 
used to obtain the image character extracted with 
DWT (Tan et al., 2014) (Tampinongkol, Herdiyeni, & 
Herliyana, 2020). 

 

𝐸 = ∑ |𝑓(𝑛)|

𝑁

𝑛=1

 

 
Where E represents energy, N represents the 
amount of data, and f (n) represents the position of 
data to n. 
 Entropy is used to measure pixel diversity 
in an image to generalize the uncertainty that 
appears in an image (Wang et al., 2015) (Widiyanto, 
Sukra, Madenda, Wardani, & Wibowo, 2018). 
 

 𝑆 = − ∫ ℎ𝑛 log2(ℎ𝑛 )
𝐺

𝑛=1
 

 

Where n is the gray level of some subband, he is the 
n-probability of the gray level, and G is the total of 
the gray level. 
 

RESULTS AND DISCUSSION 
 
Extract results using the DWT with symlet 

family wavelet, which measured information using 
energy and entropy. Energy is used to calculate the 
coefficients of LH, HL, and HH since that coefficient 
can represent the texture contained in an image. 
Whereas the LL coefficient is computed using 
entropy to measure the uniformity of the texture in 
the approximation image.  

 

 
Figure 11. Distribution of data of each pathogen 

(energy) 
 

Figure 11, The distribution of extracts 
calculated using wavelet energy, shows an overlap 
in Botrytis sp. and Culvularia sp. 2. The largest 
standard deviation is found in Botrytis sp. with 
values ranging from 925 to 647.2. Extreme outliers 
influence the value with a value of 11121374. As for 
Culvularia sp. 2., the data distribution is heavily 
piled into other classes of pathogens, which can 
cause many errors in the class Culvularia sp.2. at the 
same time, the largest outlier is shown in Culvularia 
sp.1. 

 

 
Figure 12. Distribution of data of each pathogen 

(entropy) 
 

Figure 12 shows the distribution of 
entropy values for the four pathogens and highly 
stacked data, with the highest distribution seen in 
Botrytis sp. and the largest outliers. There is also the 
main data distribution between quartile one to 
quartile third largest Culvularia sp. 1 with a range of 
1.78 605 greater than the others. 
 

CONCLUSIONS AND SUGGESTIONS 
 
Conclusion 

After DWT image feature extraction, the 
four pathogens (Botrytis sp., Culvularia sp.1, 
Culvularia sp.2, and Pestalotia sp.) exhibited 
distinctive characteristics. 

Additional findings from this study, Object 
Culvularia sp. 1 and Culvularia sp. 2. have a pixel 
intensity that is much different from the pixel 
background intensity of the image. Segmentation 
using the Otsu Thresholding method is perfect for 
such image types. The tail and the tip of Pestalotia 
sp. have a pixel intensity that is not much different 
from the pixel intensity of the background image, so 
the Otsu Thresholding method is not suitable for 
detecting the tail and the tip of Pestalotia sp. 
 
Suggestion 

The image used has an image resolution of 
under 1 megabyte. Higher quality is expected to 



P-ISSN: 2656-1743 | E-ISSN: 2656-1735 
DOI: https://doi.org/10.34288/jri.v5i1.488 

JURNAL RISET INFORMATIKA 
Vol. 5, No. 1. December 2022 

Accredited rank 3 (SINTA 3), excerpts from the decision of the Minister of RISTEK-BRIN No. 200/M/KPT/2020 

 

 
110 

 

 

capture more of the texture of a pathogen so that it 
can capture more accurate images. 

 
REFERENCES 

 
Agrios G. (2005). Plant Pathology (5th ed.). New 

York: Elsevier Academic. 
Aisah, A. R. (2014). Identifikasi dan Patogenisitas 

Cendawan Penyebab Primer Penyakit Mati 
Pucuk pada Bibit Jabon ( Anthocephalus 
cadamba (Roxb.) Miq). IPB, Bogor. 

Bangun, M. B., Herdiyeni, Y., & Herliyana, E. N. 
(2016). Morphological Feature Extraction of 
Jabon’s Leaf Seedling Pathogen using 
Microscopic Image. TELKOMNIKA 
(Telecommunication Computing Electronics 
and Control), 14(1). 
https://doi.org/10.12928/telkomnika.v14i1.
2486 

Hadi S. (2001). Masalah Dalam Perlindungan Hutan 
Terhadap Ganguan oleh Penyakit. In  Patologi 
Hutan Perkembangannya di Indonesia. Bogor: 
Fakultas Kehutanan IPB. 

Herliyana, E. N. (2013). Biodiversitas dan potensi 
cendawan di indonesia. Bogor: IPB Press. 

Herliyana, E. N., Sakbani, L., Herdiyeni, Y., & Munif, 
A. (2020). Identifikasi Cendawan Patogen 
Penyebab Penyakit pada Daun Jabon Merah 
(Anthocephalus macrophyllus (Roxb.) Havil). 
Journal of Tropical Silviculture, 11(3), 154–
162. https://doi.org/10.29244/j-
siltrop.11.3.154-162 

Larekeng, S. H., Qalbi, N., Rachmat, A., Iswanto, I., & 
Restu, M. (2022). Effect of gamma iradiated 
seeds of Jabon Merah (Neolamarckia 
macrophylla (Wall.) Bosser) to genetic 
diversity. IOP Conference Series: Earth and 
Environmental Science, 1115(1), 012027. 
https://doi.org/10.1088/1755-
1315/1115/1/012027 

Madhu, & Kumar, R. (2022). A hybrid feature 
extraction technique for content based 
medical image retrieval using segmentation 
and clustering techniques. Multimedia Tools 
and Applications, 81(6). 
https://doi.org/10.1007/s11042-022-
11901-8 

Naga Kiran D, & Kanchana V. (2019). Recognition of 
glaucoma using otsu segmentation method. 
International Journal of Research in 
Pharmaceutical Sciences, 10(3), 1988–1996. 
https://doi.org/10.26452/ijrps.v10i3.1407 

Otsu, & N. (1996). A threshold selection method 
from gray-level histograms. IEEE Trans. on 
Systems, Man and Cybernetics, 9(1), 62–66. 
Retrieved from 

https://cw.fel.cvut.cz/b201/_media/courses/
a6m33bio/otsu.pdf 

Rafael C. Gonzalez, & Woods, R. E. (2008). Digital 
Image Processing. Hoboken, New Jersey: 
Prentice Hall. 

Santosh, N. K., & Barpanda, S. S. (2020). 4. Wavelet 
applications in medical image processing. In 
Predictive Intelligence in Biomedical and 
Health Informatics (pp. 63–90). De Gruyter. 
https://doi.org/10.1515/9783110676129-
004 

Streets, R. B. (1972). The Diagnosis of Plant Diseases: 
A Field and Laboratory Manual Emphasizing 
the Most Practical Methods for Rapid 
Identification. Tucson, Arizona: University of 
Arizona Press. 

Sudarsan, B., Ji, W., Adamchuk, V., & Biswas, A. 
(2018). Characterizing soil particle sizes using 
wavelet analysis of microscope images. 
Computers and Electronics in Agriculture, 148, 
217–225. 
https://doi.org/10.1016/j.compag.2018.03.0
19 

Tampinongkol, F. F., Herdiyeni, Y., & Herliyana, E. N. 
(2020). Feature extraction of Jabon 
(Anthocephalus sp) leaf disease using discrete 
wavelet transform. TELKOMNIKA 
(Telecommunication Computing Electronics 
and Control), 18(2), 740. 
https://doi.org/10.12928/telkomnika.v18i2.
10714 

Tan, C., Wang, Y., Zhou, X., Wang, Z., Zhang, L., & Liu, 
X. (2014). An Integrated Denoising Method for 
Sensor Mixed Noises Based on Wavelet Packet 
Transform and Energy-Correlation Analysis. 
Journal of Sensors, 2014. 
https://doi.org/10.1155/2014/650891 

Wang, S., Yang, X., Zhang, Y., Phillips, P., Yang, J., & 
Yuan, T.-F. (2015). Identification of Green, 
Oolong and Black Teas in China via Wavelet 
Packet Entropy and Fuzzy Support Vector 
Machine. Entropy, 17(12). 
https://doi.org/10.3390/e17106663 

Warisno, & Dahana K. (2011). Peluang Investasi: 
Jabon Tanaman Kayu Masa Depan. Jakarta: 
Gramedia Pustaka Utama. 

Widiyanto, S., Sukra, Y., Madenda, S., Wardani, D. T., 
& Wibowo, E. P. (2018). Texture Feature 
Extraction Based On GLCM and DWT for Beef 
Tenderness Classification. 2018 Third 
International Conference on Informatics and 
Computing (ICIC), 1–4. IEEE. 
https://doi.org/10.1109/IAC.2018.8780569