COMPOSITE IMAGELET IDENTIFIER FOR ML PROCESSORS

International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 10 Issue: 08 | Aug 2023 www.irjet.net p-ISSN: 2395-0072
© 2023, IRJET | Impact Factor value: 8.226 | ISO 9001:2008 Certified Journal | Page 767
COMPOSITE IMAGELET IDENTIFIER FOR ML PROCESSORS
Ajay Kumar Naik G1, Suresh Babu B2, Srinivasan V3, Mohammed Aslam C4, Lakshmi Kiran M5
1Assistant Professor, Dept. of ECE, CBIT, Proddatur, Andhra Pradesh, India
2 Assistant Professor, Dept. of ECE, CBIT, Proddatur, Andhra Pradesh, India
3 Assistant Professor, Dept. of ECE, CBIT, Proddatur, Andhra Pradesh, India
4 Associate Professor, Dept. of ECE, CBIT, Proddatur, Andhra Pradesh, India
5 Associate Professor, Dept. of ECE, CBIT, Proddatur, Andhra Pradesh, India
---------------------------------------------------------------------***---------------------------------------------------------------------
Abstract - Image transformations have played a vital role
in capturing relevant data from resizing, conversion, edging
and pixilation strategies for better processing of explorable
data from the image lets. They have been using extensivelythe
approximation methods with finite differences used to
manipulate Edges have weight representing energy in real
time pictures captured by cameras with moderate and high
resolutions. Deployment of such applications are found in
forestry animal husbandry without spoiling the biome,
detecting animal cruelty and enhancing safety of humans
against uncontrolled fauna. AI machines of future are digital
variants of panorama and aerial image processors.
Key Words: CNN, Computer Vision in Machine Learning,
SSIM, FPGA, APR-AI-ML
1.INTRODUCTION
Graphics management tools like photoshop, fotoflexer,
amazon image, in addition to PRISM APR have built-in
addons intersecting aspect ratio of the image section that
you want to designate with n same locations that need
regeneration. SSIM [1] values are metrics in such seam tools
in multimedia but they involve manual intervention based
on requirement.EnergyEnhancementfunctionsareinvolved
in Industry toolsets like Pegasus APA, AI toolsandComputer
Vision techniques like Tensor flow, Open CV, Keras deep
learning etc. as content aware image targeting to focus on
the observer faction mainly based onDijkstra's algorithm. In
this paper a simulation of such pixilation and edge
transformation is done on real time images to comparetheir
performance on light weight devicesthatarequickerinseam
process than high density image capturing devices.
1.1 Conventional image processing methods
Before CNN are incorporated to assess whether an image
has been modified by seam carving. Though the proposed
research is not an intelligent fake image detection and
tampering in digital images [1], but utilizing the methods to
track image modifications with minimum motion pictures
instead of videos that requires either GPUs or FPGA high
density [3] chips to process the image data with limitations
in storage and retrieval compared to magnetic tapes in
traditional big data storage.
The diagram shown below contains partial computations
as part of dynamic programming in finding lowest-energy
vertical seam, for each pixel in a row submission. Shown in
Fig 1b) are the simulation results from MATLAB release-20
with both inbuilt and user-defined functions utilized to
compute the image indices for the experimental imagelisted
below.
1.2 Image manipulators
The basic processing begins with the image intensity
matrix obtained from the pixelated image, from which seam
locations are defined and manipulated with the algorithm
defined in the block diagram. Calculation of CME is for
uncompressed image is the requirement of the stature
identification algorithm that utilizes back-tracking
procedure of minimum energy along the seam path. The p-
map [1] quantization may introduce false positives as
perceptive distortions introduced or captured. The method
to differentiate between the two is discussed in the paper.
Reduction of false positives by expectation-maximization
probability techniques is out of scope of the research. The
segregation of theindividual imagesfrombigdata repository
can be later implemented for IQA with minimal degradation
in image conversion preserving the color information and
separation of identified portions of OD asthefuture relieson
cloud storage platforms mainly for AI-ML processors [5].
An approach to investigate the study on machine learning
of SEM images are helpful in magnifying the algorithmic
model to be utilized for the computer visionary of location
stature by OIM-SEAM studies [6].
2. SEM TRANSFORMATION APPLICATIONS
Image techniques are already in use in spectroscopy
(EDS), for fractography,PCB technologytestingintermetallic
distribution in solder interfaces, SSPM Seam scope
projection AUTO-XTS machines for miniature material
detection purposes. The proposed solution cited in paper is
based on mega structures identification and manipulation
utilizing the study on algorithms implemented for above
existing applications. The proposed experiments are
intermediate between existing material surveillance and
distant surveyors like SSTL S1-4 leased devices for mission

critical applications. The processing of oblique and vertical
resolutions as in GIS are beyond scope of the current
experiment. Only JPEG and SVG images are utilized in the
experiment.
Gradient magnitude, entropy, visual saliency, eye-gaze
movement are MERL seam algorithms that are compared in
the below screens with the input image. Combinatorial
optimization techniques like greedy algorithms with
variations, implemented for above existingapplications. The
proposed experiments are intermediate between existing
material surveillance and distant surveyors like SSTL S1-4
leased devices for mission critical applications. The
processing of oblique and vertical resolutions as in GIS are
beyond scope of the current experiment. The basic algorithm
uses following main equations for manipulation of seam
lines indicated in equations 1, 2 and 3.
1. Seam equation
S = [min ∑ e(I(si)))]
Energy vector
[i, j] = e [i, j] + min (M [i - 1, j], M [i, j], M [i + 1, j]);
2. Seam sn1 is defined for coordinates n 1=1,
2…, N as
sn1= {(n1, T (n1))} ∀n1|T (n1) −T (n1−1) |≤ 1
3. Accumulative cost matrix M (n1, n2) for all
possible seam connections
M(n1,n2)=e(n1,n2)+min(M(n1−1,n2−1),M(n1−1,
n2),M(n1−1,n2+1))
Image resizing with K-neighbor algorithm implementing
salient regions with parts of the background of entire
regional spatial content, is unaccountable for spatial losses
comparing boundary element method. Shown below is the
seam technique with layers for resized image output. The
decoding of images to the quantized values is achieved by
existing Sobel–Feldman mainly used for computer vision is
discussed. Though thealgorithmisalreadyimplemented asa
MATLAB function for plots, its useandfurtheradditionsmay
enhance the seam detectors. The subsections of the
algorithm involved in SEAM processing is analyzed by
breaking them into program-sublets shown in Fig 1a. The
processing involves the gradient computationthatisutilized
by distributed grabber, followed bysynthesizermayprepare
the image with scaled metrics for windowing to facilitate
seam in distributed computing systems. Thus, the machine
learning process may be fully complete with algorithmic
induction of the devices computing the multiple-oblique
seams in a PC based post processing system or an android
app in future.
Fig -1a: Device Stages
Fig -1b: Stages in Processing (Device Stages)
Fig -2a: Image Energy computation
Fig -2b: Conventional image computation
Fig -2c: Image SEAM analysis and computation
Gradient
computational
matrix
Accumulative
cost matrix
Energy
manipulator
conditions
- User
defined
Optional Sobel
[I] ʘ [Xg] [Yg] @
pixel
Grabber
Synthesizer
Windowing
SEAM facilitator

2.1 Abbreviations and Acronyms
BEM: Block Element Modifier generally used in stream
text but new for images with CAPTCHA
algorithm for extended image processors
FPGA: Field Programmable Gate Array hardware chips for
for commuter visionary devices
CNN: Convolutional Neural Network processors
3. CONCLUSIONS
Image SEM trainer-based tolls are readily available with
multi-image processing GUIs that port data from miniature
electronic devices. Such data use maximized filtering to
reduces noises that are intersecting vital informationwithin
the boundary conditionsdefined. Theseammanipulation isa
different rea where the calculations based on seam need
separate algorithm. This paper has exposed the research on
Seam utilization techniques for identifiers for vital eco
system applications. The image analyzed and extracted in
Figure 1c) clearly indicates the performance levels higher
compared to moving video image processing with high
density devices. This study may be highly useful tominimize
the compatibility issues of devices that can utilize internet
storage for their futuristic data studyacrosscloudplatforms.
ACKNOWLEDGEMENT
We like to acknowledge Principal, CBIT, Director of R&D,
CBIT and staff members of CBIT for supporting us in
technical works related to research in DIP.
REFERENCES
[1] A. Sarkar, L. Nataraj, B. S. Manjunath, “DetectionofSeam
Carving and Localization of Seam Insertions in Digital
Image”, Vision Research Laboratory University of
California, Santa Barbara.
[2] L.F.S Cieslak, K. A. Pontara da Costa, J. P. Papa, “Seam
Carving Detection Using Convolutional Neural
Networks”, IEEE SACI-12, IEEE Xplore: August 2018.
[3] E. Mishra, S. Narayan, K. Lim,” FPGA Accelerated Seam
Carving for Video”, Project in Electrical and Computer
Engineering, Carnegie Mellon University.
[4] A. Garg, A. Nayyar, A. K. Singh, “Improved seam carving
for structure preservation using efficient energy
function,” Multimedia Tools and Applications, vol. 4,
April 2022.
[5] J. Pope, M. Terwilliger, “Seam Carving for Image
Classification Privacy”,DepartmentofComputerScience
and Information Systems, University of North Alabama,
Florence, Alabama, U.S.A.
[6] P. Nguyen, R. Surya, M. Maschmann, P. Calyam, K.
Palaniappan, F. Bunyak “self-supervised orientation-
guided deep network for segmentation of carbon
nanotubes in SEM imagery”, European Conference on
Computer Vision, February 2023.
[7] Izumi Ito, “Gradient based global features for seam carving”,
EURASIP Journal on Image and Video, an. 27,September2016.
BIOGRAPHIES
Ajay Kumar Naik Guguloth is
working as Assistant Professor
with CBIT and has 9 years of work
experience and has completed his
M.Tech. in VLSI from NIT,
Surathkal, India.
Suresh Babu Byri is working as
Assistant Professor with CBIT and
has 12 years of work experience
and pursuing Ph.D. in DIP from
JNTUA, India.
SrinivasanVenugopalanisworking
as Assistant Professor with CBIT
with 11 years of work experience
and has completed his Master of
Science in PCSfromUWS,Swansea,
UK.
Mohammed Aslam Chettukrindi is
working as Associate Professor
with CBIT with 19 years of work
experience and pursuing Ph.D. in
DIP from JNTUA, India.
Lakshmi KiranMukkara isworking
as Associate Professor with CBIT
and H.O.D, Department of ECE and
has been awarded Ph.D. from YVU,
Kadapa.
SSIM: Structure Similarity Index Measure
XTS: IEEE Advanced Encryption Standard
multiprocessing
AI-ML: Artificial IntelligenceandMachineLearning

COMPOSITE IMAGELET IDENTIFIER FOR ML PROCESSORS

More Related Content

Similar to COMPOSITE IMAGELET IDENTIFIER FOR ML PROCESSORS (20)

More from IRJET Journal (20)

Recently uploaded (20)

COMPOSITE IMAGELET IDENTIFIER FOR ML PROCESSORS