Change 2D to 3D in image processing Biotechnology Текст научной статьи по специальности «Компьютерные и информационные науки»

Change 2D to 3D In Image Processing

Biotechnology

Youssef Ali Alhendawi

Abstract— The main tool for scanning is the electron microscope for bio-gauge. On the other hand, Transmission electron microscopy and Atomic-force microscopy (AFM) are increasingly used for minimum size lineaments in medical, agricultural, and bio-scientific samples. It is also noted that, some natural properties, which give its information is poor and mistakes in probability of discussion, will be high. In this paper, the researcher will tackle this problem using different technique in image processing to get more clarification, sufficient information and in depth the researcher will generate it by more methods as texture map and mosaic method. And will make the comparison between the model and the other hardware laboratory. We have got a series of images that willconstrue using the above-mentioned techniques. The filtering depends on the process of replacing devices and numerical methods that help to analyze the image and samples digitally and also get the third dimension of the first and second dimension as well as find the fourth dimension to find the properties and details of the pictures and samples that we will get. We will build code on many files, collect photos, medical and botanical samples filter images, process and convert images into data files and then work on the data file to execute programs and icons for all filters and analyzes.

Keywords—image , microscopy, bioscientific, processing .

I. INTRODUCTION

Human relies very much on our visual system (eyes & brain) to collect visual information about our surrounding. Visual information refers to images and video. In the past, we need visual information mainly for survival. Nowadays, visual information is for survival as well as communication and entertainment. In order to be able to obtain good results, we need to use the process of processing images and converting them to digital format. Image processing technology is often used as two-dimensional signals when applying specific signal processing techniques. Among today's rapidly growing technologies are its applications in various aspects of the business. Image processing forms basic research in engineering and computer science disciplines as well [1].

Image processing basically includes the following three steps:

1. Bring the image with the optical scanner or by digital imaging. 2 - Image analysis and manipulation, which includes data compression and image optimization and rely on other technologies such as satellites. 3 - The results we will get from this process. What we will benefit from image processing.

Youssef Ali Alhendawi is with the Computer Department, Prince Sattam bin Abdulaziz University of Saudi Arabia (e-mail:

[email protected])

1. Visualization - control of invisible objects. 2. Improve the image to get a clear picture. 3. to infer the images that we benefit from. 4. Identify and measure different objects in a picture. 5. Know everything that is in the picture accurately Signal processing lies in precision and discipline in engineering and mathematics that deal with analysis and processing of analog and digital signals, including image storage, image filtering, and other signal processes that we need. These signals include, for example, transmission signals, audio signals, image signals and other signals etc. Of all these signals, the area that deals with this type of signal, in which the input is an image and output, is also a picture processed in the images. As our name shows, it deals with processing on images. It can be divided into two different types: 1. Analog image processing 2. Digital image processing. Image processing is a method used to do some operations on an image, in order to get a picture that we can understand and know what is in it and to get useful results and information from this picture. It is a type of signal processing in terms of input and output image. At present, reliance on image processing has become a modern technology. It is an area of basic research and studies in computer engineering. Image processing mainly includes the following three steps:

1. Bring the picture. 2. Image manipulation and analysis. 3. Results. Because analogue and digital image processing is the two types used. The first type can therefore be used for processing analogue images of printed copies such as prints and photographs. Image analysts use different fundamentals of interpretation while using these visual techniques. The second type of digital image processing technology helps manipulate digital images using computers. The three general stages to which all types of data must be subject to the use of digital technology are pre-treatment, optimization, presentation and information extraction [2]. Procedure for Paper Submission.

II. Methods Overview of Wavelet Research

Wavelets: In this paper we introduce wavelets and the discrete wavelet transform from the classical viewpoint, based on the concept multi-resolution analysis. The Fast Wavelet Transform (FWT) allows calculating a wavelet transform in an efficient way. We will not go into all deep details, but we will limit ourselves to the topics that are important for the rest of this thesis. What are Wavelets? The main idea behind wavelet analysis is to decompose a signal f

into a basis of functions ^ : f = ^ ai ^

i

f

To have an efficient representation of the signal using a

only a few coefficients i , it is very important to use a

w w

suitable family of functions '. The functions ' should match the features of the data we want to represent. Real-world signals usually have the following features: they are both limited in time (time-limited) or (space-limited in the case of images) and limited in frequency (band-limited). Time-limited signals can be represented efficiently using a basis of block functions (Dirac delta functions for infinitesimal small blocks). But block signals are not limited in frequency. Band-limited signals can be represented efficiently using a Fourier basis, but sines and cosines are not limited in time. What we need is a compromise between the pure time-limited and band-limited basis functions, a compromise that combines the best of both worlds: wavelets ("small waves''). 3. 3. 4. Why wavelets?

One of the main features of wavelets that are important for the applications that we shall consider is their good decorrelation:

Wavelets are localized in both the space/time and scale / frequency domains. Hence they can easily detect local features in a signal. Wavelets are based on a multiresolution analysis. Wavelet decomposition allows to analyze A signal at different resolution levels (scales). Wavelets are smooth, which can be characterized by their number of vanishing moments. A function defined on the interval [a , b ] has n

j f ( x ) x'dx = 0

vanishing moments if i = 0,1.....n —1

L2 =

f : j f 2( x ) dx < J

U V = L2

and their intersection contains only the zero-function:

n V ={0}

V

In the dyadic case, i.e. when each subspace " is twice as

that belongs to one of

V f ( x )

large as " 1, a function J v 7

V.

these subspaces J has the following properties:

f (x) e V . O dilation f (2x) e V

j+i

f (x)eV0 O translation f (x+1)eV0

.....(2.1)

...(2.2)

If we can find a function ^ ( X ) G Vo such that the set of

functions consisting of x) and its integer translates x-k) }ez

form a basis for the space Vo ,we call it a scaling function or father function. for the other subspaces V j ( with j* 0 )

we define:

(x) = 2j/>(2j x-k)

3. 4 Wavelet Functions

V. V. ΠV. ,

Because the subspaces j are nested: j j

V. , V. W we can decompose J in J and 1 , the orthogonal

V ^ V+1 :

complement of J in

Vj © wj =

The direct sum of the subspaces ' ' j is equal to

W ^ Vj .

^ L2:

for •>•>••••> the higher the number of vanishing moments, the better smooth signals can be approximated in a wavelet basis. Furthermore there exist fast (O(n)) and stable (wavelets can be orthogonal or biorthogonal) algorithms to calculate the discrete wavelet transform and its inverse.

3. 3. 5. Wavelets and Multi-Resolution Analysis. Classically, discrete wavelets are introduced in the concept of a multi-resolution analysis. We introduce this concept first.

3. 3. 5. 1. Multi-Resolution Analysis t2

Consider the space L , the vector space of square integrable functions in R :

1 — -+- cu

U Vj = © W. = L2

j=—J

V.

This means that 7 is a "coarse-resolution''

V W

representation of 7+1 , while 7 carries the "high-

V V

resolution" difference information between 7+1 and 7 .

Y ( x ) G W

If we can find a function v 7 0 that obeys the

translation property 3(.2), i.e.

Y(x)e W0 ^ translation x + 1)e W0

^ ( x )

and such that the set of functions consisting of and

{^(x-k) }

its integer translates 1 ik eZ form a basis for

W

the space 0 , we call it a wavelet function or mother

T2

In a multi-resolution analysis [34] we decompose L in Vj

nested subspaces J

...c V-2 c V-i c V0 c Vi cV2 c...

T2

such that the closure of their union is L ,

function. for the other subspaces

W ( with j * 0 )

we

define: W>■> (») = 2"* W(2-k)

3.4.1. The Fast Wavelet Transform (FWT)

V W V

Because both 0 and 0 are subspaces of 1 : V0 c V1 and W0 c V1 ,

(d( x ) W ( x ) we can express ^v ' and v 7 in terms of the basis

functions of V1 Due to the multi-resolution analysis, these

relations are also valid between

3. 5. Steepest descent algorithm ( SDA)

—j

—j

The gradient is a vector operator denoted ^ f = gad (f ) the gradient is given by:

V :

The relation

O ( Ämin) = f ( Pk + hmin Sk ) shows

.(2.3)

df A 1 a f A 1 af A

V/ ( x, j, z ) =--- x +--- y +--—z

hl d x h2 d y h3 d z

The direction ofV f is the orientation in which the

|V f I

directional derivative has the largest value and 1 1 is the

value of that directional derivative. Furthermore, ifV f ^ 0 , then the gradient is perpendicular to the level curve ( Xo , y o ) if z = f ( x, y )

through 0 the level surface through

and perpendicular to

( x0 , ^0 z 0 ) if f ( x, y, z ) = 0

We can generalized the equation 4.3 for " k where k = 1,2,3,........., n

Vf (x) =j8f ( -) 8f ( -) df ( -) + +df (x)

d x1

d x.

d —

d x„

We applying this function on file of data for the piping

image off( x , x2, x3 ), recall that the gradient vector in (4.4) points locally in the direction of great rate of increase

off ( x ) . Hence V f ( x ) points locally in the direction

of greatest decrease f ( x )' Start at the point Po and search

p

along the line through o in the direction

•s0 =-v f ( p0)d-V f ( p0)||' •„ • .

o o o you will arrive at a

point p1 , where a local minimum occurs when the point x

is constrained to lie on the line x Pl + vS() . Since partial derivatives are accessible, the minimization process can be executed using either the quadratic or cubic approximation method.

Next we compute - V f ( pi ) and move in the search j • • =—v f ( p1)A\-v f ( p^ll. ...

direction 1 H you will

come to p2 , where a local minimum occurs when x is

constrained to lie on the line x = Pl + v •1. Iterative will produce a sequence, { pk =0, of points with the property

f ( x0 ) ^ f ( xi) ^^ f ( xk ) ^... .f

limp^ = p then f (p)

kwill be a local minimum f (x). 3. 5. 1. Outline of the steepest descent algorithm (SDA)

Suppose that pk has been obtained.

Step 1 : Evaluate the gradient vector V f ( Pk ) .

Step 2: Compute the search direction

4 = —V f (Pk )/|| —V f (Pk )||-

Step 3 :Perform a single parameter minimization of

® (v) = f (pk + v^k )on the interval [ 0,b ],

v=h

where b is large. This will produce a value mm where a

that this is a minimum for k along

f (Pk ).

the search line X Pk + VSk .

Step 4: Construct the next pointPk+1 = + ^min Sk . Step 5: Perform the termination test for minimization, as:

f ( Pk )and

Are the function value f (Pk+i ) Sufficiently close and the distance

| pk+1 - pk 11 smaU enough? Step 6: Repeat the process.

■■(2.4)

Fig. 3.1. This figure shows the method algorithms

III. Filter

Implementation

We implemented our algorithm for two-dimensional filter using Visual Fortran and the figures have been plotted using Origin. The filtering looks of high quality since it seems to recover the original sine wave with the add noise totally removed. For two-dimensional, any of the normally deal with image files with extensions. (*.bmp) or (*.tiff) are first changed to one with extension. (*.dat). Thus each image is a data matrix. Practically, Steepest Descent Method ( Grid Model).

3.6.2 The implementation of the algorithm: After seeing the up figures we find that the algorithms is the best and reduction the costs for processing, its depends on the programming using Visual Fortran, Surfer. From the statistical analysis report we get the mean, median, mode, the linear regression equation to the pixels for the matrix for the data files, getting the Sum of Squares ( SS ) and Mean Square ( MS ). After the analysis we get the distribution for the three images as below:

local minimum for

O ( v )

Intensity of color

Fig.3.6.2.1. a. Original image

Fig. 3.6.2.2 b. The surface 3D for the original image

piiiptaöi WwÈtS '

Fig. 3

6.2.3. c. The Contour lines for the original image

. -- !

: ...

\ i

i

- ' r

. / -

- v • ^

t t

-

\ \ t

» *

{ ?

' " ; -

J.KU t

t _

t ? ».

. ; • '* ** ■

t >t

Fig. 3.6.2.4 d. The SDA and the direction for the original image

Distance

Fig. 3.6.2.5 e. The 2D, the relation between the Distance (pixels) and Gray value for the original image

Statistical analysis

This statistical analysis for original image a: Univariate Statistics

Table 3.6.3.1 for image a: Statistics

# X Y Z

Median: 0.5 0.5 0.42

Median Abs. Deviation: 0.25 0.25 0.11

Mean: 0.499 0.499 0.418

Standard Deviation: 0.288 0.289 0.160

Variance: 0.083 0.083 0.025

Coef. Of Variation: 0.57 0.57 0.38

Coef. Of Skewness: 0.001 0.001 -0.100

Planar Regression: Z = AX+B Y+C Fitted Parameters

Table 3.6.3.2 for imag ?e a: Fitted Parameters

# A B C

Parameter Value: 0.032 0.14 0.32

Standard Error: 0.001 0.001 0.0007

This statistical analysis for original image b: Univariate Statistics

Summary: In this paper deals with the extraction of 3D geometry from single still image and by the transformation on the intensity of color we can get the depth on the all images by using OpenGL The problem is decomposed into a number of tasks, each task is associated with a specific geometric group. Existing techniques have been implemented and combined to form a relatively easy algorithm, which is encountered in its source code in the OpenGL library for the texture mapping models. We can applied this method for many fields, it is easily to use successfully our method for modeling, laying out of the scene, and rendering to finding the depth from the one image as already implemented. From this paper we introduce method for converting the image formats to digital images. And using this method for samples in

Nanomaterials. However, this method can be applied to the image patterns in different ways. Giving different quality of image in each graph. We use the transformation method also. This method can be applied in the all Nanomaterials, for finding the depth by using the transformation on the intensity of color using OpenGL.

IV. Conclusion

The thesis have four models the first model is using the image processing converting model from image extinction to data file after that using the relation between pixels for more images and making the comparisons and statistical analysis. The second model is steepest descent algorithm (SDA) and wavelet transformation model. The third model is using texture map under OpenGL library on Visual C++ for finding the depth or the 3D from one still image. The model presented the extraction of 3D geometry from single still image. The problem is decomposed into a number of tasks; each task is associated with a specific geometric group. Existing techniques have been implemented and combined to form a relatively easy algorithm, which is encountered in its source code in the OpenGL library. We can applied this method for many fields, as in nuclear filed for finding and restoration the damage parts of image of building and detecting the radiation from the cracked piping. Feeder pipes in the pressurized heavy water reactor It is easily to use successfully our method for modeling, laying out of the scene, and rendering to finding the depth from the one image as already implemented. An understanding of this issue helps in planning the scene capture. Each depth image provides a sample of the 3D world from its location. Obviously, the sampling has to be dense near the areas of fine scene structure. The quality of rendered views is likely to suffer when using depth maps that are quite far. This is an issue that requires careful study. These properties make the depth image representation suitable for IBR. The representation could be used even for synthetic models as the rendering requirements for a particular view could be lower. We are currently exploring the following aspects of the representation. Blending functions should be defined so that the influence of every view tapers off smoothly. This will eliminate the artificial .edges. in rendered view when the captured images differ in color or brightness. We are studying the compression of the depth maps and the texture images together, taking advantage of the properties and constraints of the geometry of the input views. Finally our thesis, we introduce method for converting the image formats to digital images. And using this method for samples in Nanomaterials. However, this method can be applied to the image patterns in different ways. Giving different quality of image in each graph. We use the transformation method also. This method can be applied in the all Nanomaterials. We have a statistical result as mean, median, standard deviation, variance and sum of squares (SS) and mean square (MS) are shown in tables above. Then the steepest descent algorithm is a useful tool for digital image processing because it can be applied iteratively. Find 3D from 2D by more models as Warping model, Texture mapping.

References

[1] Klipiteostus: Gholamreza Anbarjafari "Digital Image Processing" ,

http://www.uttv.ee/naita?id=20081

http://youtu.be/FPNGPHkXybo?list=UU-

ETlxdihAaw8Pn6_Zz10lg.

[2] I.a.ismail, M.A.Ashabrawy, Using Image's Processing Methods in Bio-Technology, Int. J. Open Problems Compt. Math., Vol. 2, No. 2, 2009.

[3] M.A. Ashabrawy, N. M. A. Ayad, "Using Grid Model for Nano measurements images", International Journal of Latest Research in Science and Technology, pp.228-233, 2014.

[4] M. A. Ashabrawy, "Using Gradient Model to Compare Between Treatment Samples and Non-Treatment Samples", IEEE Xplore Digital Library, London, pp.1440-1449, 2015.

[5] Erik Smistad, Thomas L. Falch, Mohammadmehdi Bozorgi, Anne C. Elster and Frank Lindseth "Medical Image Segmentation for Improved Surgical Navigation ", Medical Image Analysis, volume 20, February 2015, pages 1-18. Elsevier B.V.

[6]Ender GunesYegin, KorkutYegin & Osman CavitOzdogan"Digital image analysis in liver fibrosis : basic requirements and clinical implementations ",Pages 653-660 | Received 31 Jan 2016, Accepted 20 Apr 2016, Published online: 12 May 2016. To link to this article: http://dx.doi.org/10.1080/13102818.2016.1181989.

[7] M. Schwier, T. Chitiboi, T. Hulnhagen, H.K. Hahn "Automated spine and vertebrae detection in CT images using object-based image analysis ", Institute for Medical Image Computing, Fraunhofer MEVIS, Bremen, Germany. First published: 14 August 2013.

[8] Matthew S. Simpson, Dina Demner-Fushman, Sameer K. Antani, George R. Thoma "Multimodal biomedical image indexing and retrieval using descriptive text and global feature mapping", June 2014, Volume 17, Issue 3, pp 229-264.

[9]AminTermehYousefi (ChECAIKohza),Samiraagheri (Nanotech nology & Catalysis Research Centre (NANOCAT), NahrizulAdib "Integration of biosensors based on microfluidic: a review", (2015) "Integration of biosensors based on microfluidic: a review", Sensor Review, Vol. 35 Issue: 2, pp. 190-199, doi: 10.1108/SR-09-2014-697

http://dx.doi.org/10.1108/SR-09-2014-697

[10] Denis Vadlja ,Martin Koller Email author, Mario Novak ,Gerhart Braunegg, Predrag Horvat "Footprint area analysis of binary imaged Cupriavidusnectar cells to study PHB production at balanced, transient, and limited growth conditions in a cascade process ",ApplMicrobiol Biotechnol. 2016 Dec;100(23): 1006510080. Epub 2016 Oct 3.

[11] Esmael Hamuda, , Martin Glavin, Edward Jones"A survey of image processing techniques for plant extraction and segmentation in the field",Computers and Electronics in Agriculture archive Volume 125 Issue C, July 2016.

[12] Sachin D. Khirade, A. B. Patil "Plant Disease Detection Using Image Processing", 2015 International Conference on Computing Communication Control and Automation.

[13] A. Camargo, J.S. Smith, "An image-processing based algorithm to automatically identify plant disease visual symptoms", Jun 08, 2016.